Pre-dispersal predation of Taraxacum officinale (dandelion) seed

Authors


A. Honek (tel. +420 233 022269; fax +420 233 311591; e-mail honek@vurv.cz).

Summary

  • 1Pre-dispersal predation of seeds of Asteraceae has been studied in species where ripening seed is present on plants for long periods but rarely in those where seed maturation is ephemeral and density of consumers is therefore unlikely to keep pace with changes in seed availability.
  • 2We therefore followed predation of dandelion (Taraxacum officinale) seeds by the larvae of Glocianus punctiger and Olibrus bicolor and predicted that both abundance of seed consumers and seed damage will be indirectly proportional to inflorescence availability, and that overall seed damage will be less than in species with longer lived inflorescences.
  • 3We counted the number of dandelion capitula m−2, number of larvae capitulum−1 and percentage of damaged seeds at 10 sites, where the flowering time and densities of dandelions differed. The counts were made in 2002 and 2003, at half-weekly (April–May) or weekly (June–August) intervals.
  • 4Abundance of both species of consumer varied among dandelion patches, and with change in availability of dandelion capitula. Numbers of larvae capitulum−1 were high early and late, when few capitula were available, but decreased at the time of peak flowering when there were many capitula. Production of ready-to-pupate larvae m−2 of a species at a site was similar in successive years, but values for O. bicolor and G. punctiger were not correlated.
  • 5Seed damage paralleled the abundance of consumer larvae, with early and late flowers suffering most. A linear relationship correcting for lost seeds, predicted 30% damage when there were five larvae capitulum−1, below levels reported for species of Asteraceae with persistent inflorescences.
  • 6The intensity of pre-dispersal seed predation is directly proportional to the abundance of seed consumer species and indirectly proportional to the availability of maturing capitula. Consequently, in species with ephemeral, synchronized flowering, where seed is available for only a short time, the majority of plants lose only a small proportion to predators. Although those flowering earlier or later than the peak will suffer a much higher risk, the low overall level of damage is unlikely to influence population biology.

Introduction

As one of the most important factors influencing seed mortality, pre-dispersal predation may even influence the population biology of species (Bevill et al. 1999). Maturing seeds are eaten because of their abundant energy and nutrient content, typically by small, rapidly developing insect species that move over only short distances (Crawley 1997). The seed may be visibly damaged by chewing species, such as coleopteran or dipteran larvae, although the consequences of attack by sucking insects are often unapparent (Gould & Sweet 2000). Predator occurrence is limited by seed availability, which changes in space and time. Opportunities for seed predation vary according to the size and structure of fruits and seeds, the length of time that the seed remains on the plant and the degree of chemical protection. Seed consumers are therefore favoured by the simultaneous production of large numbers of seeds (Silvertown 1980; Kelly 1994), a strategy that also protects the seed producer by spreading the risk of predation (Kelly et al. 1992).

The inflorescences of Asteraceae usually contain many comparatively large seeds and are frequently damaged by abundant chewing seed predators (Diptera, Campbell et al. 2002; Coleoptera, Louda et al. 1997; and Lepidoptera, Louda et al. 1997; Cummings et al. 1999) whose abundance is correlated with the size of the capitulum (Fenner et al. 2002). The average proportion of damaged seeds reported has usually been high (e.g. Briese 2000), possibly because ripening capitula of the species studied persisted on the plants for several weeks and therefore allowed development of the seed consumer population. Overall damage is the product of the number of seeds eaten per infested capitulum (frequently 100%) and the percentage of capitula infested (sometimes also high). The performance of seed predators may, however, be much more limited in species whose flowering and seed production are ephemeral. Because of the short period available, each larva is able to consume very few seeds and adults are unable to exploit the available resources effectively. We therefore propose that, in short-flowering species, the abundance of consumers will be negatively correlated with the availability of resources, because consumers are unable to adjust to the rapidly changing availability of host-plant inflorescences. Consequently, we predict that seed damage will be indirectly proportional to seed availability and that average seed damage will be less than in long-flowering species, because of the limited time available for seed predation. To test these hypotheses, we studied seed predation in dandelion, a species in which small patches of plants typically produce large quantities of seeds over short periods of time.

Dandelion, Taraxacum officinale G.H. Weber ex Wiggers, is an aggregate of many apomictic species whose differences, however, might be unimportant for the seed consumers. This globally distributed perennial colonizes a range of grassy areas, including parks, gardens, pastures, orchards, roadsides and ruderal areas (Stewart-Wade et al. 2002). In central Europe, flowering and seed production mainly occur during April–May and September–October, although in some places reproduction may occur throughout the vegetative season. Reproductive effort depends on site and plant size (Welham & Setter 1998), with a plant producing up to 20 capitula (diameter varies between 4 and 15 mm, seed number from 40 to 380 seeds and total seed mass from 0.01 to 0.27 g, Honek & Martinkova 2002). Individual seed size varies both between and within the capitula (Welham & Setter 1998; Tweney & Mogie 1999). In the field (late April to early May), an inflorescence persists for 23.1 ± 0.5 days (mean ± SE), of which 9.1 ± 0.7 days is allocated to stalk extension, 2.7 ± 0.2 to flowering, 9.6 ± 0.7 to seed maturation and 1.7 ± 0.1 to seed dispersal (A. Honek & Z. Martinkova, unpublished data). Although seed maturation makes up 42% of the total duration of the inflorescence, the time available for development of seed consumers is short because of the low temperatures prevailing during maturation. Unlike the temperatures of flowers (Orueta 2002), those of closed dandelion inflorescences parallel air temperature (A. Honek & Z. Martinkova, unpublished data), which for April–May is 10.3 ± 0.3 °C (20 years average).

The dominant seed consumers in central Europe are two species of beetle (Coleoptera), a weevil Glocianus punctiger (Gyllenhal) (Curculionidae) and Olibrus bicolor (Fabricius) (Phalacridae), whose larvae complete their development within dandelion capitula (Campagna & Rapparini 2002; Honek et al. 2005). Other seed consumers, Cnephasia oxyacanthana (Herrich-Schäfer) (Lepidoptera: Tortricidae) and Ensina sonchi (Linné) (Diptera: Tephritidae), are rare. Feeding on dandelion seeds by G. punctiger (Radde 1974; Lohse 1983) and O. bicolor (Heyer, cited in Reitter 1912; Vogt 1967) has been known for many years. The eggs of G. punctiger are laid in the hollows of dandelion stalks, at the beginning of stalk extension. The apodous larva moves to the underside of the thalamus, tunnelling through the receptacle to feed on the maturing seeds, before pupating in the soil following seed dispersal. Larval development is completed during the pre-dispersal stage of an inflorescence and dandelion capitula with ripe seeds are therefore mostly populated by old larvae about to pupate (Martinkova & Honek 2003; A. Honek & Z. Martinkova, unpublished data). The life history of North American populations of G. punctiger was studied by McAvoy et al. (1983) but some of their results are inconsistent with our data. The basal temperature for development of O. bicolor is higher (13.5 °C) than the value we recorded for G. punctiger (6.3 °C), and its larvae are therefore unable to complete development within one capitulum and, after seed dispersal, migrate to other maturing inflorescences. Pre-dispersal dandelion capitula therefore contain a mixture of young and old larvae, which also pupate in the soil. Young larvae of both species eat holes in the seeds, whereas old larvae consume parts of or whole seeds. Because of the complete destruction of some seeds it is difficult to estimate total consumption (Honek & Martinkova 2002). The larvae also may cut the plumes and prevent seed dispersal.

In order to test our hypotheses (above) we carried out a 2-year study in which we investigated local and annual variation in dandelion seed production and abundance of O. bicolor and G. punctiger. We also assessed the effects of abundance of beetles and dandelion capitula on seed damage, and the influence of seed damage on mortality and germination.

Materials and methods

study area

The study was carried out in 2002 and 2003, in a 0.18-km2 area at Prague-Ruzyne (50°06′ N, 14°15′ E, altitude 340 m a.s.l.), which is a mosaic of experimental fields, orchards, ornamental gardens and ruderal areas separated by field tracks and ditches. Sites consisted of small stands of dandelion plants that had established naturally within the swards and were separated from one another by areas of different vegetation or buildings. The 10 sites, which were separated by at least 35 m, differed in size, exposure to sun, frequency of cutting, intensity of trampling and sward quality (Table 1). Sward quality was evaluated by mean total above-ground plant biomass (estimated on 30 May 2003, i.e. after peak seed production), of 5 × 0.5 m2 randomly placed plots at each site, which were cut, dried and weighed. Cutting at this time influenced dandelion flowering very little because any inflorescences were replaced within c. 1 week.

Table 1.  Site quality and flowering of dandelion. ARE = area of site (m2), SUN = insolation (O = open, S = shaded), CUT = sward cut (U = 1–2 times per season, C = > 2 times per season), TRA = trampling (T = frequently, N = rarely), SWR = sward quality (P = poor site, < 200 g above-ground dry mass m−2 on 30 May, F = fertile site, > 200 g above-ground dry mass m−2 on 30 May), START = beginning of flowering, DUR = duration (days) of flowering at a particular site (including late summer and autumn when seed consumers were absent), DEN = maximum numbers of flowering capitula m−2 (always attained during the spring flowering peak)
SiteSite qualityDandelion flowering
20022003
ARESUNCUTTRASWRSTARTDURDENSTARTDURDEN
  • *

    Area size in 2002 was 100 m2, in 2003 the area was increased to include a marginal population of dandelion flowering throughout the whole season.

  • Shaded for only a part of the day.

1120SUNF26 April 2534.229 April 2236.6
2 90SUNF21 April 3655.529 April 3962.9
3 90OUNP15 April 3228.718 April 2929.9
4 80SUNF26 April 2819.929 April 2518.1
5200OCNP21 April10554.125 April11650.9
6113OCTP15 April10143.729 April 9940.4
7250*OCNP 8 April 4311.815 April12613.5
8120SCNP 8 April 3911.115 April 2913.8
9114OUTP 8 April13035.5 9 April14635.7
10221OCTF26 April 2564.725 April 3670.1

abundance of seed consumers

Seed consumers were counted in mature dandelion capitula (1–3 days before seed dispersal) collected at 3- to 4-day (April–May) or weekly intervals (June–October). At each site, two replicate samples of 15 (2002) or three samples of 10 (2003) randomly sampled inflorescences were cut off just below the capitulum, put into 0.5-L plastic bottles covered with nylon mesh and stored at room temperature (24–26 °C). Within 24 hours of collecting, each sample was put on a sieve (0.5 cm mesh diameter) placed over a Petri dish (25 cm diameter), the capitula dissected and the larvae counted. The apodous larvae of G. punctiger are easy to distinguish from the campodeiform larvae of O. bicolor. Larvae of both species were divided into ‘large’ (last instar larvae ready to pupate) and ‘small’ (larvae of earlier instars not capable of pupating). The instars, which can be identified by head capsula size (McAvoy et al. 1983; Honek & Martinkova, unpublished data), were identified visually (by AH) because measuring all larvae was impracticable.

dandelion seed production

Flowering capitula were counted at each site in 50 randomly placed 0.25-m2 quadrats at weekly intervals, from mid-April to early October We present data for the period when G. punctiger and O. bicolor were present. As there was no vertebrate grazing or selective harvesting of maturing capitula, the mean number of inflorescences m−2 was assumed to be equal to the number of ‘mature’ capitula (containing seeds 1–3 days before dispersal) 1 week later. Sites 5, 6, 7 and 9 were cut in June, well after the spring peak of flowering, but as predators were present, predator ‘production’ at this time was corrected for consumers lost in the cut. However, the effect of cutting on total seed consumer ‘production’ was very small because of the low density of capitula at this time. During April–May, when seed predators were sampled more frequently than flower production (at 3- to 4-day vs. weekly intervals), inflorescence numbers on intervening dates were calculated as the mean of the two adjacent values.

seed damage and germination

During May–June 2003, samples of mature seeds were collected at weekly intervals. At each site, seeds were collected from 30 capitula with open involucra, but prior to seed dispersal. The seeds were dried under laboratory conditions (26 °C, 40% relative humidity) for 3 days, then stored dry at −15 °C until germination tests were done. For each site and date, five samples of c. 100 seeds were taken at random and sorted into heavily damaged (> one-third seed volume missing), slightly damaged (missing < one-third seed volume including small scars) and undamaged seeds. Ten collections, covering the range of sampling dates and seed damage were selected in order to test germination of heavily damaged, slightly damaged and undamaged seeds (three replicates of 50 seeds in each damage category). If damaged seeds in a given collection were scarce, the number of seeds per replicate (50) was decreased (minimum n = 20) but the number of replicates (three) was maintained. Germination of all seeds was tested simultaneously with each group of 50 (or fewer) placed on a dense, slow-filtering paper (Filtrak®, Bärenstein, Germany) moistened by 5 mL tap water in a glass Petri dish of 10 cm diameter, and kept at constant 25 °C and a 4 hours light : 20 hours dark photoperiod. Germinating seeds were counted at 2-day intervals until no germination was observed in any dish for 4 days.

data processing

For each site and date, the average number of capitula m−2, and average number of large and small larvae of each species capitulum−1 were calculated. For each species, the density (m−2) of individuals ready to pupate at seed dispersal was calculated as (number of large larvae capitulum−1 on a particular day) × (number of flowering capitula m−2 half a week previously). The time shift allowed for the period needed for seed maturation. For each site, these numbers were summed over the whole period of seed maturation to give an estimate of ‘production’.

To establish the relationship between predator abundance and seed damage, the proportion of seeds that were damaged and dispersed on a particular day was plotted against the average number of larvae of both species present at the site 3–4 days earlier. The relationship was fitted by a second order polynomial model

image(eqn 1 )

where y is the proportion of damaged seeds and x is the number of larvae capitulum−1. To select the optimal model the data were fitted by linear and curvilinear regression and the proportion of explained variance calculated. The polynomial model was selected because the quadratic term significantly (P < 0.05) increased the proportion of explained variance (Crawley 1993). The proportion of seeds affected by predators (damaged and completely eaten) was estimated by fitting a tangent y = a0 + a1x at y = 0. The value for x (0.237) was obtained from equation 1.

To study the general trends in dandelion capitula and seed predator abundance the data for all sites were synchronized and/or standardized. Dandelions at particular sites started to flower on different dates. For synchronization, the start of flowering at each site was set as day 1 and subsequent dates were assigned the number of Julian days that had elapsed since that date. Average number of capitula m−2, number of beetles capitulum−1 and proportion of damaged seeds varied greatly among sites. For standardization, the maximum value at each site was assigned the value 1 and other values were expressed as a proportion of this value.

The effect of site quality on species abundance was tested by one way-anovas with annual ‘production’ of G. punctiger or O. bicolor, at each of the 10 sites tested against one of four parameters of site quality (exposure to sun, frequency of cutting, intensity of trampling and sward quality). Between-site differences in production of seed predators were tested by nested anova with production (2002 and 2003) as response variable and species as factors, nested within sites. Differences in germination (standardized) of seeds with different degrees of damage were evaluated by analysis of covariance ancova with percentage germination on successive days as response variable, seed damage as a factor (no damage and slightly damaged vs. heavy damage) and day of germination as a covariate (using data for days 3–15).

Arithmetic means (± SE) are presented throughout the paper. All calculations and model fitting were done using Excel, and anova was calculated using Statistica for Windows (StatSoft 1994).

Results

dandelion flowering

Dandelion patches differed in date of onset and duration of flowering (Table 1). The start of flowering at particular sites differed by 18 (2002) to 20 (2003) days, but the spring peak number of flowering capitula was always observed 7–14 days (mean 10.2 ± 0.94 days) after the onset (Fig. 1a). The number of flowering capitula decreased within 1 week of this peak and thereafter flowering was sporadic. Flowering occurred over 3–5 weeks (22–39 days in 2003) at six sites and several months (99–146 days) at four sites. At two sites numbers of flowering capitula increased in late summer when seed predators were no longer present. Maximum number of flowering capitula m−2 differed between sites but were similar at the same site in 2002 and 2003 (Table 1). Because the pattern of flowering was typical for particular sites, starting date (R2 = 0.707, P < 0.005), duration of flowering (R2 = 0.977, P < 0.001) and maximum abundance (R2 = 0.970, P < 0.001) are highly correlated between 2002 and 2003.

Figure 1.

Seasonal change in the incidence of dandelion flowering and abundance of seed predators at 10 sites (synchronized data for 2003): (a) standardized numbers of dandelion capitula m−2 (relative to a maximum of one at each site). (b) Numbers of Glocianus punctiger larvae capitulum−1. (c) Numbers of Olibrus bicolor larvae capitulum−1. For each plot the data start on day 1, for which the mean is the first filled circle (± SE); subsequent filled circles are means calculated for successive observations over 3- or 4-day periods.

seed predation

The occurrence of seed consumers was determined by the availability of mature dandelion seeds, which paralleled the flowering activity with a 1-week delay. With few exceptions (sites 4 and 5 in 2003), larvae of both species were found at the start of seed production. The duration of their presence at a site (Table 2) was limited by length of the dandelion seed-production period and presence of both G. punctiger (R2 = 0.99, P < 0.001) and O. bicolor (R2 = 0.94, P < 0.001) was therefore correlated with the duration of flowering at a particular site. The duration of the two species at particular sites (Fig. 2) was highly correlated (R2 = 0.94, P < 0.001) but O. bicolor was present for a significantly shorter period than G. punctiger (nested anova: MS = 2945.2, F10,20 = 7.807, P < 0.001).

Table 2.  Populations of Glocianus punctiger and Olibrus bicolor at sites 1–10, in 2002 and 2003. D = number of days when species were present; P = production, number of ready to pupate larvae m−2 produced over the season; Abundance = average abundance (± SE) over the early period (days 1–15) and late period (days > 15) of species occurrence
SiteG. punctigerO. bicolor
DPAbundanceDPAbundance
Days 1–15Days > 15Days 1–15Days > 15
2002
12118.00.2 ± 0.040.5 ± 0.2021 7.60.2 ± 0.090.3 ± 0.00
22125.80.1 ± 0.030.4 ± 0.0821183.92.1 ± 0.283.2 ± 0.43
33079.01.2 ± 0.330.8 ± 0.3530 49.90.7 ± 0.090.7 ± 0.09
4251.90.1 ± 0.060.1 ± 0.0818  7.90.1 ± 0.040.4 ± 0.19
510573.70.2 ± 0.051.3 ± 0.2472118.00.9 ± 0.230.7 ± 0.20
610122.20.1 ± 0.021.1 ± 0.176871.51.9 ± 0.710.9 ± 0.19
73021.40.5 ± 0.140.4 ± 0.143020.62.1 ± 0.671.2 ± 0.09
83613.60.1 ± 0.050.3 ± 0.06363.70.1 ± 0.050.2 ± 0.06
9123117.50.8 ± 0.191.5 ± 0.258126.20.2 ± 0.010.5 ± 0.11
102524.60.1 ± 0.050.4 ± 0.152577.21.2 ± 0.621.0 ± 0.15
Mean  0.3 ± 0.120.7 ± 0.15  0.9 ± 0.270.9 ± 0.28
2003
12216.00.4 ± 0.101.1 ± 0.322221.10.5 ± 0.100.2 ± 0.04
23643.40.2 ± 0.051.8 ± 0.4136146.42.6 ± 90.972.0 ± 0.38
32571.70.7 ± 0.170.9 ± 0.142552.21.2 ± 0.360.5 ± 0.07
42511.90.2 ± 0.060.2 ± 0.03150.30.1 ± 0.010.0 ± 0.00
599121.80.6 ± 0.050.7 ± 0.168068.40.7 ± 0.130.6 ± 0.15
6865.20.0 ± 0.020.8 ± 0.215247.60.7 ± 0.211.1 ± 0.35
710636.10.6 ± 0.111.1 ± 0.406658.52.0 ± 0.240.9 ± 0.15
82921.20.4 ± 0.050.7 ± 0.10250.90.1 ± 0.030.1 ± 0.02
913783.60.9 ± 0.080.9 ± 0.378863.31.4 ± 0.030.3 ± 0.13
102929.90.1 ± 0.040.8 ± 0.1629133.11.1 ± 0.230.8 ± 0.20
Mean  0.4 ± 0.070.9 ± 0.13  1.0 ± 0.260.7 ± 0.19
Figure 2.

The number of days for which Glocianus punctiger and Olibrus bicolor larvae are present at the 10 sites, data for 2003. Regression: a0 = 8.894, a1 = 0.588, R2 = 0.939, P < 0.001.

Abundance at different sites showed similar patterns for the two species. Initial abundance (small and large larvae pooled) decreased within c. 10 days of the start of seed production (when dandelion capitula were most abundant) but increased again from c. day 20, more obviously in G. punctiger (Fig. 1b) than in O. bicolor (Fig. 1c). Because larval abundance varied over time, an average value was calculated for the period before (day 1–15 of seed production) and after (day > 15) the peak of dandelion flowering. In particular years, the differences between sites varied by a factor of 9–12 for G. punctiger and 21–26 for O. bicolor (Table 2). Average abundance of G. punctiger was less before the peak because, although only sporadic dandelion capitula were found after the peak, larvae were still abundant in the few capitula found. For O. bicolor, the difference between the flowering periods was small because the overall abundance of this species decreased during the season.

The time when the cumulative production of larvae m−2 increased steeply coincided with the time of peak ripening of dandelion capitula. Before and after this period cumulative production changed slowly because ripening capitula were scarce and therefore the numbers of larvae m−2 were low (Fig. 3). The change in production of larvae ready to pupate was thus determined by dandelion rather than predator abundance and consequently occurred at the same time for both predators. The time to 50% of total production was highly correlated (R2 = 0.735, P < 0.005) between the two species (Fig. 4a).

Figure 3.

Production of individuals at the 10 sites. Numbers of ready to pupate larvae m−2 accumulated over the period of dandelion flowering, data for 2003. (a) Glocianus punctiger. (b) Olibrus bicolor.

Figure 4.

Correlation between characteristics of the Olibrus bicolor and Glocianus punctiger populations at the 10 sites, data for 2003. (a) Time T50 (Julian days from January 1) to cumulative 50% production of individuals m−2. Data for site 6 not included in calculation of the correlation because the low abundance of G. punctiger made the estimate of T50 unreliable. Regression: a0 =−62.42, a1 = 1.522, R2 = 0.735, P < 0.01. (b) Total production m−2 over the season. Regression: a0 = 43.90, a1 = 0.347, R2 = 0.069, NS.

The total production of the two seed consumers differed between sites by a factor of 62 (2002) and 35 (2003) in G. punctiger, and 50 (2002) and 488 (2003) in O. bicolor (Fig. 3, Table 2). Total production at a particular site was determined by numbers of larvae capitulum−1, which varied more than the numbers of dandelion capitula m−2. As the abundance of the predator species differed, production of the two species at each site was therefore not correlated (R2 = 0.069, NS) (Fig. 4b). Dandelion flowering (Table 1) and average abundance of larvae (Table 2) were similar in both years and production in 2002 and 2003 was therefore highly correlated for both species (Fig. 5). There was no correlation between site quality and the production of G. punctiger or O. bicolor adults (one-way anovas for particular species and factors: F1,8 = 0.0097–2.449, NS). There was a significant relationship between maximum density of dandelion capitula and production of O. bicolor (R2 = 0.715, P < 0.05), but not of G. punctiger (R2 = 0.056, NS). Duration of flowering was not significantly correlated with the production of either species (R2 = 0.015–0.249, NS) because number of dandelion capitula m−2 in summer was low and this period contributed little to production.

Figure 5.

Correlation between the production of seed predators (numbers of ready to pupate larvae m−2) at the 10 sites in 2002 and 2003. (a) Glocianus punctiger, regression: a0 = 11.11, a1 = 0.829, R2 = 0.682, P < 0.005. (b) Olibrus bicolor, regression: a0 = 20.52, a1 = 0.682, R2 = 0.654, P < 0.005.

seed damage

Variation in beetle abundance led to variation in damage level between seed collections. At seed dispersal, 0.0–12.3% of the seeds in a sample were slightly damaged and 0.0–12.3% were heavily damaged (combined: 0.0–22.0% damaged). Seed damage was proportional to the abundance of larvae of both species present 3 days previously (Fig. 6). This relationship was best described by a second order polynomial (a0 = −0.00154, a1 = 0.065, a2 = −0.0054, R2 = 0.835, P < 0.001).

Figure 6.

Proportion of seeds damaged vs. number of larvae capitulum−1 (Glocianus punctiger plus Olibrus bicolor) present half a week before seed dispersal, data for 2003. Heavy line = second order polynomial fitted to the data, regression: a0 = −0.0154, a1 = 0.065, a2 = −0.0054, R2 = 0.835, P < 0.001. Dashed line = model allowing for completely eaten seed, a0 = −0.015, a1 = 0.0623.

Because seed predator abundance was determined by the availability of dandelion capitula, seed damage decreased following the peak of flowering, i.e. at the time of maximum seed production (Fig. 7a,b). The ln of the average proportion of seeds damaged and average abundance of seed predators in the previous 5 days are significantly correlated (R2 = 0.908, P < 0.005) (Fig. 7c).

Figure 7.

Dandelion capitula m−2 and seed damage, synchronized data for 2003. Variation over time of (a) dandelion flowering (synchronized) and (b) seed damage (proportion of seeds damaged). Filled circles represent average values ( SE). (c) Relationship between averages of seed damage and dandelion flowering. Regression: a0 = ln 0.2958, a1 = −0.6538, R2 = 0.908, P < 0.005.

Seed damage significantly affected percentage germination (Fig. 8a). Slight damage decreased average germination to c. one-third of the control. The heavily damaged seeds (> one-third eaten) rarely germinated and, on average, percentage germination was only one-tenth of that of slightly damaged seeds. However, seed damage did not affect the germination rate (Fig. 8b), as that of undamaged and slightly damaged seeds was virtually the same, and that of heavily damaged seed, although retarded, was not significantly so (ancova: MSeffect = 0.036, F1,14 = 1.210, NS). The hypocotyl of damaged seeds usually appeared first and then the radicle, if it appeared at all.

Figure 8.

The effect of seed damage on germination. (a) Percentage of seeds that germinated (± SE). (b) Cumulative proportions of germinated seeds (± SE). Categories of damage: undamaged, slightly damaged (< one-third seed eaten), and heavily damaged (> one-third seed eaten).

Discussion

seed predators

Abundance (number of individuals capitulum−1) and production (number of larvae ready to pupate m−2) of both species varied among sites during the season. Although the distances between sites were small, the differences in abundance and production were significant and persisted over the 2 years. It is unlikely that this variation was caused by differences in local abundance of the predators as adults of both species overwinter far away from the dandelion patches (J. Strejcek, personal communication). Although Anthocorid adults and larvae (Heteroptera), and Syrphid (Diptera), Chrysopid (Neuroptera), Staphylinid and Cantharid (Coleoptera) larvae are present in ripening dandelion capitula (Honek et al. 2005), and could cause mortality of egg and larval stages, they are rare, and therefore unlikely to have a large effect on the abundance of the seed consumers, and would not consume the seed predator species selectively. Differences are therefore presumably due to differential ovipositon: G. punctiger females might respond to host-plant quality, as shown for other ceutorhynchine weevils (Buchi 1996; McCaffrey et al. 1999) in which host-plant quality is evaluated by olfactory cues (Saringer 1976). However, we cannot generalize these results as there are no data on odour perception by G. punctiger or on dandelion kairomones.

The different habitat preferences of the two seed predators might be related to ways in which they exploit dandelion capitula (Martinkova & Honek 2003). G. punctiger larvae are each limited to a single inflorescence, while those of O. bicolor are unable to complete their development in one capitulum and migrate between or among capitula. Olibrus bicolor females might therefore select sites where the density of plant cover is low and soil surface texture is smooth, and therefore more suitable for larval migration. The production of both seed predators and the quality of stands did not differ between years.

The pattern of change in numbers of larvae capitulum−1 of both species accords with our prediction that availability of capitula and seed consumer abundance would be negatively correlated. Abundance of both species decreased during the peak period of capitula ripening. Consumer populations, despite their increase in size, were ‘diluted’ by the great abundance of capitula, and numbers capitulum−1 temporarily decreased.

seed damage

Calculating the seed consumption by particular species was difficult because the larvae usually occurred together. Their numbers and the proportion of seeds damaged in particular capitula were only loosely correlated (Z. Martinkova & A. Honek, unpublished data), perhaps because migrating larvae of O. bicolor might have arrived in a capitulum just before it was sampled. However, there was a relationship at the population level between predator numbers and seed damage. The increase in the proportion of damaged seeds with average number of larvae is better fitted by a second order polynomial than a linear relationship, possibly because of the complete consumption and therefore loss of some seeds. To determine the numbers of seeds lost we used the tangent, to the zero damage point of the second order polynomial fitted to the experimental data. This linear model (a0 = −0.015, a1 = 0.0623) allowed for the lost seeds (Fig. 6) and predicts that 29.7% of the seeds will be completely eaten or damaged when there are five larvae per capitulum. As this is above the larval density in more than 95% of our samples, this may be the upper limit to seed damage under natural conditions. McAvoy et al. (1983) estimated that a larva of G. punctiger consumes 19 seeds during its development, consistent with five larvae consuming one-third of the seeds in a large capitulum containing 250–300 seeds.

We predicted that seed damage should be negatively proportional to seed availability. The change in numbers of flowering capitula appeared to determine predator numbers capitulum−1. Seed damage decreased because larval populations were diluted at the time of peak seed production and the risk of seed predation decreased. Plants that flower immediately after the peak are at greatest risk of predation as numbers of capitula decrease dramatically and the seed predators are still abundant.

A maximum proportion of 30% of seeds consumed is consistent with our prediction that the short time available for predation limits the extent of seed damage in dandelion, despite occasional high abundances of predator larvae. This contrasts with higher proportions of seed in the capitula of other Asteraceae destroyed by weevil (Scott 1996; Turner et al. 1996; Louda et al. 1997; Cummings et al. 1999; Briese 2000; Louda & O’Brien 2002; Maron et al. 2002; Vickery 2002) or tephritid larvae (Vaishampayan et al. 1970; Lamp & McCarthy 1982; Steck 1984; Forsyth & Watson 1985; Straw 1985), with some cases approaching 100%, despite lower larval densities (Lalonde & Roitberg 1994). The time available for predation is often longer, between 13 and 40 days (Fenner et al. 2002). Dandelion has a maximum of c. 9 days and time available for seed predation may be even shorter because larvae eat thalamus and involucrum tissues for part of the seed ripening period, switching to seeds when they become more nutrient-rich than the green parts of the capitulum.

Acknowledgements

We thank Professor A. F. G. Dixon for critical reading of the manuscript, helpful suggestions and improving the English, S. Pekar for assistance in calculations, and Mrs H. Uhlirova and Mrs L. Kreslova for excellent technical assistance. The work was supported by grant no. 521/03/0171 of the Grant Agency of the Czech Republic and grants 0002700601 (ZM) and 0002700603 (AH) of the Ministry of Agriculture of the Czech Republic.

Ancillary