The effect of slug grazing on vegetation development and plant species diversity in an experimental grassland

Authors

  • H. BUSCHMANN,

    Corresponding author
    1. Geobotanical Institute, Swiss Federal Institute of Technology, Zürichbergstrasse 38, CH-8044 Zurich, Switzerland
      *Author to whom correspondence should be addressed. E-mail: hbuschm1@gwdg.de
    Search for more papers by this author
    • Present address: Albrecht-von-Haller-Institute for Plant Sciences, University of Göttingen, Grisebachstr. 1a, D-37077 Göttingen, Germany.

  • M. KELLER,

    1. Geobotanical Institute, Swiss Federal Institute of Technology, Zürichbergstrasse 38, CH-8044 Zurich, Switzerland
    Search for more papers by this author
  • N. PORRET,

    1. Geobotanical Institute, Swiss Federal Institute of Technology, Zürichbergstrasse 38, CH-8044 Zurich, Switzerland
    Search for more papers by this author
  • H. DIETZ,

    1. Geobotanical Institute, Swiss Federal Institute of Technology, Zürichbergstrasse 38, CH-8044 Zurich, Switzerland
    Search for more papers by this author
  • P. J. EDWARDS

    1. Geobotanical Institute, Swiss Federal Institute of Technology, Zürichbergstrasse 38, CH-8044 Zurich, Switzerland
    Search for more papers by this author

*Author to whom correspondence should be addressed. E-mail: hbuschm1@gwdg.de

Summary

  • 1Generalist herbivores such as slugs have the potential not only to reduce plant density and biomass, but also to alter species diversity within vegetation. Their impact on species diversity may be either negative, if they concentrate feeding on less abundant plant species, or positive, if they feed on the most abundant species.
  • 2This study investigated the influence of slugs on plant species diversity in experimental swards produced by sowing a Lolium perenne/Trifolium repens seed mixture in field plots with a large seed bank of mainly arable species. Half of the plots were grazed by Arion lusitanicus Mabille. Plant cover, above-ground biomass and number of plant species were measured over a 3-year period.
  • 3Vegetation cover increased in the control plots from 50% in the first year to 90% in the third year. Cover was significantly lower in the slug plots in the first year (>22%), while there were only small differences between treatments in the third year. Slugs reduced total above-ground biomass by >25% in both the first and third years.
  • 4Slugs had a negative impact on plant species diversity in the first year, particularly by reducing forb species. In contrast, plant species diversity after 3 years was higher in the slug plots than in the controls, because of the higher number of forb species. Under slug grazing, the biomass and cover of annual and palatable species were reduced, but not the numbers of these species.
  • 5Our results suggest that slugs can have a significant effect on plant species diversity in plant communities, but that the direction of the effect changes during the course of succession. In the earliest stages, when most species are present as seedlings or juveniles, slug grazing leads to reduced species diversity because favoured species are eliminated. In closed vegetation, in which competitive interactions are important, slugs may reduce the dominance of the more competitive species and thus provide gaps in which plants can establish from seed. As a consequence, slugs tend to cause an increase in plant species diversity, and may also reduce the rate of successional change by promoting the persistence of annual species.

Introduction

By feeding on plant material, invertebrate herbivores may not only reduce the biomass and productivity of vegetation, but also alter plant species composition. The processes by which they do so are complex, and many factors may play a role (Rees & Brown 1992; Crawley 1989, 1997). For example, if the main herbivores select only one or a few plant species, then the abundance of these species is likely to be reduced, and they may even be eliminated from the vegetation. However, this does not necessarily lead to a loss of plant species diversity; if the selected plants are highly competitive species, herbivory could promote species diversity by allowing other, less vigorous species to persist in the vegetation (Brown & Gange 1989; Bach 1994; Olff & Ritchie 1998 and references therein).

Slugs are among the most important herbivores in low herbaceous vegetation in the cool-temperate zone (Rees & Brown 1992; Hulme 1996; Rodríguez & Brown 1998), and can influence both the biomass and species composition of plant communities (Oliveira Silva 1992; Hanley et al. 1995a; Hulme 1996; Bruelheide & Scheidel 1999). Although many slug species are more-or-less polyphagous, they usually show distinct preferences for certain plant species and may avoid feeding on others (South 1992; Briner & Frank 1998; Frank & Friedli 1999). Because plants are most vulnerable to slug damage as seedlings (Hanley et al. 1995a; Frank 1998a, 1998b; Hanley 1998 and references therein), slugs are most likely to influence the plant species composition of vegetation in the earliest stages of secondary succession, when most plants are present as seedlings (cf. Dirzo 1980; Edwards & Gillman 1987; Hanley et al. 1995a, 1995b; Wardle & Barker 1997). In contrast, established plants are more likely to survive the effects of slug grazing, although their vigour and fitness may be reduced (Ehrlén 1995).

Some studies have demonstrated how slugs can reduce species diversity and increase the rate of secondary plant succession by selecting more palatable species, commonly annual and early successional perennial species (Cates & Orians 1975; Reader & Southwood 1981; Fraser & Grime 1999). Other work has shown that food selection may also be affected by the relative abundance of plant species. When slugs prefer to feed on less abundant species (anti-apostatic selection), as demonstrated by Cottam (1985), they tend to reduce the plant species diversity of vegetation (Bruelheide & Scheidel 1999; Frank 2003). In contrast, species diversity can be enhanced if slug grazing reduces the abundance of common species (Olff & Ritchie 1998).

Most experiments to investigate the impact of slug grazing on plant species composition have been of short duration, lasting from a few weeks up to 2 years, and have focused on the seedling phase rather than on established vegetation (Hulme 1994, 1996; Hanley et al. 1995a, 1995b, 1996; Frank 1998a, 1998b). One exception is found in the work of Fenner et al. (1999), who observed effects of slugs on the species composition of a herbaceous community due to the differential survival of seedlings in vegetation gaps created 5 years earlier by Hanley et al. (1996). However, our knowledge of the influence of slugs on the course of secondary succession remains limited.

In this study we investigated the impact of slug grazing on biomass and species composition during the first 3 years of a succession on arable land. We hypothesized that slugs have different impacts at the beginning of a succession, when most plants are present as seedlings, and after closed vegetation has established. Specifically, we expected slugs to have a negative effect on plant species diversity at the beginning of the succession due to selection of palatable seedlings, and to have a positive effect on plant diversity in the later stages of vegetation as a result of their reducing the competitive ability of the more abundant species. The questions posed in this study were: (i) How do slugs influence vegetation development and the diversity of plant species establishing from the seed bank? (ii) Does the influence of slug herbivory change during the course of succession? (iii) Are contrasting ecological groups of plants, annual vs perennial, and highly palatable vs less palatable plant species, affected differently by slug herbivory?

We investigated these questions by sowing a clover (Trifolium repens)–ryegrass (Lolium perenne) sward in a soil with a large seed bank of mainly ruderal species. This mixture was chosen because these species are commonly sown as part of an arable rotation, and it is known that the development of the sward leads to reduced slug populations (Frank 1998a, 1998b). As an experimental herbivore we chose the slug Arion lusitanicus Mabille, a common agricultural pest in central Europe (Reischütz 1986; Frank 1998a, 1998b). The design of the experiment allowed us to investigate how slugs influence the establishment of weed species in a sward initially dominated by the sown species.

Materials and methods

The experiment was established in April 1997 on former arable land at a level, open site (105 m2) on base-rich, loamy soil in the experimental garden at ETH Hönggerberg, Zurich, Switzerland (523 m a.s.l.) and was run until September 1999. In order to reduce the patchiness of the natural seed bank, the uppermost soil (≈15 cm) was removed, mixed thoroughly and returned to the site. Fourteen plots were established, each surrounded by commercial slug-proof frames (2 × 2 m2, height 34 cm, depth in soil 10 cm). The frames were set out in a regular grid with a spacing of 75 cm between neighbouring frames (Fig. 1).

Figure 1.

Arrangement of the experimental area and design of experimental plots (slug frames) with subplots used to measure plant cover in 1997. 0, slugs absent; 1, slugs added.

The plant varieties sown in the experiment were Lolium perenne var. Bastion (13 kg ha−1) and Trifolium repens var. Ladino Regal (4 kg ha−1); nomenclature of plant species follows Oberdorfer (1990). Lolium perenne is reported to be rather unpalatable to slugs (Dirzo 1980), while the chosen variety of T. repens has a low content of cyanogenic glycosides and is therefore attractive to slugs (Dirzo & Harper 1982; Horrill & Richards 1986).

The seeds were sown on 14 April 1997. There was vigorous regeneration from fragments of various clonal species, notably of Vicia spp., Ranunculus spp., Achillea millefolium, Taraxacum officinale agg. and Cirsium arvense, and during the first 2 months it was necessary to remove these species by weeding. Thereafter it was no longer possible to distinguish between vegetative regrowth and seedlings (which were the subject of this study), so weeding was discontinued.

The slugs for the experiment were collected from seminatural vegetation in the surroundings of the experimental area at Hönggerberg. Five weeks after sowing (12 May 1997) half of the plots, chosen at random, received 22 slugs in the first year and 10 slugs in consecutive years, while slugs were excluded from the rest of the plots. The numbers of slugs per plot were chosen to represent high natural slug densities (Frank 1998a), thus conforming to the recommendations of Hanley et al. (2003) for mollusc-feeding experiments. Four wooden shelters were placed in each plot, 10 cm from each side, to protect the slugs during drought (Keller et al. 1999; Fig. 1). From March to November the number of slugs per plot was monitored on a weekly basis. Missing slugs were replaced and surplus individuals and other mollusc species were removed. A molluscicide (metaldehyde pellets) was applied twice to prevent slugs from entering the slug-exclosure control plots. In accordance with the usual management in agricultural clover–ryegrass fields in Switzerland, all plots were mown twice in the first and second years, and three times in the third year (using a motor scythe).

data collection

The early development of the vegetation was monitored closely by taking photographs of two subplots per plot (30 × 45 cm2) every second day from 24 May to 9 June 1997. These subplots were located 30 cm from the fence in the centre of the long side of each half-plot (Fig. 1). Using image analysis (Dietz & Steinlein 1996), the vegetation cover was calculated from the photographs. In the third year, when the vegetation cover was nearly 100%, this technique was no longer feasible; thus the percentage cover of each plant species was estimated visually.

In the first year the above-ground yield of each plant species was measured in eight randomly placed subplots (20 × 20 cm2). A slightly different procedure was used in the third year, when vegetation composition in all plots was recorded by cutting samples within a randomly placed ring of 0·125 m2 (four samples per plot). Because of the quantities involved, it was no longer feasible to separate the harvested material into individual species, and it was therefore divided into five categories (herbs, legumes, grasses, litter, mosses). All material was oven-dried at 80 °C and weighed. The number of species present in the plots was recorded each year.

Information about the palatability of plant species was obtained from the literature (Grime et al. 1968; Duval 1971; Jennings & Barkham 1975; Dirzo 1980; Rathcke 1985; Mølgaard 1986; Cook et al. 1996; Wardle & Barker 1997; Briner & Frank 1998; Scheidel & Bruelheide 1999; Peters et al. 2000; Schädler et al. 2003). Life-history information (annual and biennial vs perennial) was obtained from Oberdorfer (1990).

data analysis

Cumulative dry mass of the different harvests in one year was used for data analysis (two harvests in 1997; three in 1999). The Shannon–Wiener index and Shannon evenness (Magurran 1996) were calculated based on biomass data for the species in the first year, and based on the cover values obtained for the third year (Table 1). The following equations were used:

Table 1.  Vegetation cover-abundance code adapted from Braun-Blanquet (cf. Westhoff & van der Maarel 1973 and references therein) used for plant cover analysis in 1999. For clonal plants, ramets were counted as individuals
CategoryCover (%)Abundance (individual)Weights
R 1 or 2 0·1
+<1>2 0·5
1 1–5or >10 3
2 5–10or >50 7·5
310–20or >10015
420–30 25
530–50 40
image

where p is the biomass of species (first year) or cover (third year) and S is the number of species.

jmp 5·0 (SAS Institute Inc. 2002) was used for all statistical analyses; values of P < 0·05 were accepted as significant.

The t-tests were used separately for each year to test for treatment differences in diversity indices, total dry mass and total vegetation cover. If the data deviated significantly from a normal distribution (Shapiro–Wilks test), the Wilcoxon test was used.

Log-linear analysis was used to test for differences in the number of plant species per life form (annual vs perennial) and per palatability category (palatable vs unpalatable) between years and between slug treatments. Separate analyses were run for life form and palatability.

Repeated-measures manova was used to investigate possible shifts in the dominance of annual (including biennial) vs perennial species, and palatable vs unpalatable species, over time in relation to slug treatments. Separate analyses were run for life form and palatability. Information was available on the dry mass of individual species in 1997, while for 1999 only cover data were available. To investigate changes in species abundance between the two years, the species data for each year and treatment were ranked, and mean rank values compared. As total cover in 1999 was correlated with total biomass for the same year (Spearman's rank correlation, R = 0·63, P < 0·001), we assume the ranking of these two parameters was a valid way to compare the data for the 2 years. In addition, Röttgermann et al. (2000) showed that biomass of several species in open herbaceous vegetation increases linearly with vegetation cover.

Results

development of vegetation

The two sown species, L. perenne and T. repens, germinated and established well. As intended, large numbers of seedlings of weed species also established from the seed bank: among the most abundant were Stellaria media, Euphorbia helioscopia, Cichorium intybus and Silene alba.

Even at the first recording on 24 May 1997, when the sward was 7 weeks old and slugs had been present for just 2 weeks, there was a clear reduction in plant cover in the plots with slugs (grazed 12%, control 19%; Fig. 2; Table 2). This difference was also evident at the last recording on 9 June 1997, when the total plant cover in grazed plots was 40% compared with 52% in controls (P < 0·05; Table 2). Mean above-ground plant yield in 1997 (summed data for two harvests) was reduced by slug grazing, being 130 g m−2 in control plots and around 25% lower in grazed plots (P < 0·01; Table 2). Yields of the two sown species, L. perenne and T. repens, were reduced under grazing by 22 and 36%, respectively. In both control and grazed plots, weed species derived from the seed bank accounted for ≈50% of the first year yield (data not shown).

Figure 2.

Development of vegetation cover in 1997 and cover abundance of vegetation at harvest time in 1999 (plot means ± SE; n = 7 replicate plots). Open circles, slug plots; filled circles, control plots.

Table 2.  Percentage cover and above-ground yield of vegetation, species number, species diversity and evenness in plots with and without slug grazing (mean values ± SE)
ParameterDate19971999
Control plotsSlug plotsControl plotsSlug plots
  1. Significance of differences between treatments (t-test, except for total yield in 1997 for which the Wilcoxon test was used): ***, P < 0·001; **, P < 0·01; *, P < 0·05.

CoverMay19 ± 2*12 ± 3*  
Jun52 ± 3*40 ± 6*90 ± 583 ± 3
Mean dry mass of yield per harvest (g m−2)Aug65 ± 4**49 ± 6**93 ± 3***66 ± 4***
Total dry mass of yield per year (g m−2)Aug130 ± 3**96 ± 10**278 ± 5***198 ± 11***
Number of species (total)Aug 50544464
Number of species (mean)Aug29 ± 1*24 ± 3*25 ± 1**33 ± 1**
Shannon–Wiener indexAug1·5 ± 0·1***1·2 ± 0·2***2·3 ± 0·1*2·5 ± 0·1*
Shannon evennessAug0·05 ± 0·010·05 ± 0·010·71 ± 0·010·71 ± 0·02

In the third year, the mean percentage cover of vegetation was 83% in the slug plots and 90% in the control plots (Fig. 2; Table 2). The sown species accounted for only a low proportion of the total cover, but there were marked differences in the relative proportions of these species between treatments: thus L. perenne had a cover of 10·7% in the control plots but only 3·9% in the grazed plots; the corresponding values for T. repens were 7·3 and 2·2%. The yield recorded at the end of the experiment (August 1999, summed data for three harvests) was 28% lower in the grazed plots than in control plots (198 and 278 g m−2, respectively, P < 0·001; Table 2). Weed species excluding legumes made up 46% of total yield in control plots and 59% in slug plots (data not shown); the equivalent values for grasses (mainly L. perenne) were 36 and 31%, respectively, and 18 and 10% for leguminous species (of which T. repens contributed <50%).

species diversity

A total of 92 plant species were recorded during the course of the experiment. Many of these had low biomass and cover values, were represented by only a few individuals, and occurred in only 1 year (21 species only in 1997; 34 species only in 1999). Species observed only in the first year were annual species (with one exception), while species observed only in the third year were mostly perennials. Twenty-one (control plots) and 32 (slug plots) species were present in both years. The mean number of species per plot ranged from 27 in 1997 to 29 in 1999, but the differences between years were not significant (t-test: df = 26, t = −1·4, P = 0·17).

The presence of slugs significantly affected the mean number of species per plot, although the direction of the effect changed with time. In the first 2 years, mean number of species per plot was lower in the slug plots than in controls (15 and 10% lower in years 1 and 2, respectively; Table 2). However, in the third year species numbers were 23% higher in the slug plots. These differences were also reflected by the mean value of the Shannon–Wiener diversity index, which was lower in slug plots than in controls in 1997, but higher in 1999 (Table 2). As Shannon evenness was similar in the 2 years in both treatment and control plots (Table 2), it is clear that the differences in the Shannon–Wiener index were caused mainly by the number of species.

The proportion of annual species declined strongly between 1997 and 1999 (Fig. 3; Table 3; P < 0·001), especially in grazed plots. During the same period the proportion of unpalatable species also increased; this trend was slightly greater in the slug plots than in the controls, but the difference was not significant. The occurrence of species with contrasting life-history traits was investigated in more detail by calculating the mean dominance ranks for the paired categories annual/perennial and palatable/unpalatable. This analysis confirmed the declining prominence of annual species with time (Fig. 4; Table 4; P < 0·001). It also revealed that annual species had a lower ranking in grazed than in control plots, particularly in 1999 [42 ± 3 (slugs), 35 ± 2 (control); P < 0·05], while the perennials showed no differences in rank between treatments [37 ± 2 (slugs), 39 ± 3 (control); P = 0·52]. Similar results were obtained for palatability categories. While the mean dominance rank of palatable species decreased over the course of the experiment, the mean rank of unpalatable species increased (P < 0·001). Palatable species had a higher ranking in both years in the control plots, while unpalatable species had a higher ranking in the slug plots in 1999 (Fig. 4; Table. 4; P < 0·05).

Figure 3.

Changes in total species number over time of annual vs perennial plant species and palatable vs unpalatable plant species (information collected from several studies, see Materials and methods) in 1997 and 1999. Open circles, annual; filled circles, perennial; open squares, palatable; filled squares, unpalatable.

Table 3.  Results of log-linear analysis; treatment factors are slug density (slug vs control plots) and year (1997 vs 1999)
Source of variationLife formPalatability
  1. Shown are χ2 values of main and interaction effects on number of species per life strategy (annuals vs perennials) and palatability (palatable vs unpalatable) (df = 1 in each case; other explanations as in Table 3).

Slug density 1·4   0·2
Year21***   4·4*
Year × slug density 0·1<0·1
Figure 4.

Changes in mean species rank in biomass and cover abundance over time for annual vs perennial and palatable vs unpalatable plant species (information collected from several studies, see Materials and methods). Low rank expresses high biomass or high cover. Open circles, annual (n = 45); filled circles, perennial (n = 47); open squares, palatable (n = 26); filled squares, unpalatable (n = 26).

Table 4.  Results of repeated-measures manova of possible shifts in dominance of annual (including biennial) vs perennial species and palatable vs unpalatable species
Source of variationLife formPalatability
OverallIndividualOverallIndividual
AnnualPerennialHighLow
  1. Treatment factors are slug density (slug vs control plots) and years (1997 vs 1999). Shown are F values of main and interaction effects on species rank in biomass or cover, respectively. Tests were run separately for life form and palatability (df = 2 for overall tests; df = 1 for individual tests; other explanations as in Table 3).

Slug density 3·2* 5·2* 0·7 5·9** 4·9* 5·2*
Year87***23*** 28***18***16***15***
Year × slug density 5·5** 5·3*<0·1 1·0 0·6 1·6

Discussion

The results demonstrate that slugs can have a major impact on the development of vegetation by reducing the above-ground biomass and production of vegetation, and thus delaying the development of complete plant cover. Reduced plant cover due to slug grazing was evident as early as 2 weeks after the onset of the treatments, and was still detectable in the third year. Perhaps more surprising was the fact that slugs reduced yield by around 25%, not only in the first year when the plants were mainly small, but also in the third year. In this respect our results contrast with those of Hulme (1996) and Frank (2003), who found differences in plant density due to slug grazing, but no effects on plant biomass.

Slug grazing had a considerable effect on plant species composition. At the beginning of the experiment the vegetation was dominated by the sown species, and the yield of these was reduced by slugs. In the case of L. perenne this reduction was approximately in proportion to the overall yield reduction, while T. repens was affected disproportionately. In the third year the negative impact of grazing on the two species was even greater, and they were much less abundant in grazed plots than in controls (3% cover of L. perenne, 2% of T. repens). Although T. repens tends to be rather unpalatable to slugs (Dirzo & Harper 1982), we deliberately chose a variety with low cyanide content, which may explain why this species was strongly reduced by grazing. Lolium perenne is also known to be rather unpalatable to slugs (Dirzo 1980), and the reduction in its abundance in the grazing plots could be because slugs ‘felled’ seedlings and young shoots without feeding on them, as observed by Hatto & Harper (1969); Dirzo & Harper (1980).

In the first year, the mean species number and diversity were lower in the slug plots than in the controls. Species almost absent from the slug plots included Matricaria inodora, Malva silvestris and Geranium pusillum. The loss of certain plant species in these plots is probably a direct consequence of the feeding preferences of slugs, and possibly also of differences between seedlings in their tolerance of grazing damage. With the high proportion of bare ground at this early stage in succession, competitive interactions between plants were probably relatively unimportant in limiting some species.

In the third year, slug grazing promoted plant species richness and diversity. At this stage the biomass of vegetation was considerably greater and the plant cover was nearly 100%; thus competition probably became a more significant factor than in the first year (Schädler et al. 2003). It was noticeable that three of the most common perennial species, Daucus carota, T. repens and L. perenne, were vigorous in the control plots but substantially reduced in the slug plots. This observation suggests that an important effect of slug grazing was to reduce the vigour of the abundant species and allow less competitive species to persist. We observed that slug grazing also helped to maintain a more heterogeneous environment with a higher proportion of small gaps (Dirzo 1980; Cottam 1985), which may have been important in allowing weedy species to establish from seed (Crawley 1989, 1997). In addition, the reduced quantity of litter in the slug plots (data not shown) may have provided more favourable conditions for seed germination (Wilby & Brown 2001).

In contrast to our results, there was a higher species richness in plots without slugs in the early old-field succession studied by Wilby & Brown (2001). However, in their experiment the vegetation reached a maximum cover after 3 years of 70%, and thus more closely resembled the conditions during the first year of our experiment. The study of Hanley et al. (1996) confirms our finding that slugs are able to enhance plant species diversity. However, their results were for seedling communities in small gaps in grassland; from our experiments we would predict less diversity at this stage. This contradiction may be because the seed community of grasslands is very different from that of the ruderal vegetation we studied, as grasslands contain more mid-successional species.

The analysis of species composition in terms of ecological groups of plants revealed relatively minor changes due to slug grazing. Although the palatable species were reduced in biomass by slugs, we found neither fewer palatable species in slug plots than in control plots, nor more unpalatable species. This discrepancy could reflect a methodological problem, because we used data from earlier leaf-disc assays to determine the palatability of species. Results of these palatability assays should be used cautiously as they may not reflect palatability of the species under field conditions (Hanley 1998; Fenner et al. 1999). There was also no significant reduction of annual plant species in the grazed plots, although these tend to be more palatable than perennial species (Grime et al. 1968; Cates & Orians 1975; Reader & Southwood 1981). Between 1997 and 1999, annual species that were dominant in the control plots in 1999 were reduced in dominance in the grazed plots. This could explain why, although the overall biomass of annual species in the grazed plots was reduced, the number of species tended to be higher in 1999 than in 1997. Apparently herbivory by slugs had different influences at the population and community levels. At the population level, herbivory had negative effects for most individuals, leading to a reduced dominance of annual species (perhaps because annuals tend to be less able than perennial species to compensate for leaf loss; Grime 2001). This effect may have been stronger in 1999 than in 1997 because the annuals were less tolerant of grazing under the greater pressure of competition in 1999. In contrast, at the community level, annuals may profit from slug herbivory if grazing reduces local extinction of competitively subordinate species and provides favourable conditions for germination and establishment (Olff & Ritchie 1998).

Acknowledgements

We are grateful to T. Pfau for donating the slug fences and to Johannes Kollmann, Sabine Güsewell, Mick E. Hanley, Michael Fenner and an anonymous reviewer for providing helpful comments on the manuscript. We thank Bettina R. Kahlert for her assistance with the field work. The study was supported financially by ETH Zurich.

Ancillary