was reported from Ipomopsis aggregata (Pursh) V. Grant. (Polemoniaceae) by Hainsworth et al. (1984). In the case of B. alpina, however, this does not lead to a ...
Oecologia 9 Springer-Verlag1989
Oecologia (1989) 81 : 181 185
Predispersal seed predation in Bartsia alpina Ulf Molau t, Bente Eriksen 1, and Jette Teilmann Knudsen 2
1 Department of Systematic Botany, University of G6teborg, Carl Skottsbergs Gata 22, S-413 19 G6teborg, Sweden 2 Department of Chemical Ecology, Box 33031, S-400 33 G6teborg, Sweden
Summary. A northern Swedish population of Bartsia alpina,
an arctic-alpine perennial herb, was found to suffer high levels of predispersal seed predation by larvae of two insect species, both specialists on rhinanthoid Scrophulariaceae hosts. The primary predator is Aethes deutschiana (Lepidoptera-Tortricidae), the host of which was previously unknown. The other predator is Gimnomera dorsata (DipteraScatophagidae), which is basically a Pedicularis specialist. Both predators are attacked by larvae of Seambus brevicornis (Hymenoptera-Parasitica-Ichneumonidae). Total predation pressure was more or less constant during 1985-1987, but in 1988 the level was doubled, the possible reasons of which are discussed. Large inflorescences of B. alpina suffer significantly higher predation pressures than small ones. It is shown that predation is most intense in the middle of the infiorescences. The same floral nodes are known to produce more selfed seeds than distal and basal nodes. Seed predation in B. alpina thus results in an increased proportion of outcrossed seeds entering the seed pool. Selection pressures on host plant and predator fauna are discussed. Key words: Bartsia alpina - Seed predators - Parasitoid - Relative node position - Selection pressure
Many plant species suffer high levels of predispersal seed predation, especially by more or less host-specific insects (Janzen 1971a). Host-specific seed predators often enforce strong selection pressure on a plant's timing of its blooming season (Breedlove and Ehrlich 1968; Zimmerman 1980; Augspurger 1981). There are several records in the literature regarding plants' avoidance of too high a level of seed predation by escape in time or space (Janzen 1971 a, b; Beattie etal. 1973; Platt etal. 1974; Harper 1977; Augspurger 1981). However, all the previous studies of seed predation concern tropical or temperate plants, and the possibilities for subarctic plants, such as B. alpina, to escape predation in time is much more restricted due to the short vegetation period. Furthermore, no studies known to us deal with the effects of predator-induced selection on the genetical variation among the seeds actually entering the seed pool. During a study of the mating system and pollen-mediated gene flow in a population of Bartsia alpina L. (Scrophulariaceae) in N Sweden (Molau et al. 1989), a large proOffprint requests to." U. Molau
portion of the uncaged control plants turned out to be attacked by seed predators. Larvae of two different predators were abundant in the Bartsia capsules, and an inventory of predator damage in remaining infructescences from previous years showed that it is regularly exposed to high predation pressure. Since a regular predispersal seed predation of this magnitude should influence the gene flow within populations of the host, we realized that a study focusing on this subject was essential for a complete understanding of the reproductive ecology of B. alpina. Furthermore, apart from a brief comment by Sryrinki (1939) that seed predators were found in B. alpina capsules in N Finland, the entire host-predator complex was unknown. No Fennoscandian insect was known to utilize B. alpina as host plant. The larvae found in 1986 could not be determined to the species level, but an examination of the material revealed that there was one microlepidopteran (Tortricidae) and one fly (Diptera) species in the predator sample, and that there was also a parasitic wasp (Hymenoptera-Parasitica). In the following, we examine the nature of predispersal seed predation and the consequences for the host species in a population of Bartsia alpina. We hypothesize that the seed predators are coevolved with their host plant species, so that, when habitats are relatively stable, as in this case, a steady state of counteracting selection pressures is maintained. We address the following questions: (1) How extensive is the seed predation and which insects are involved? (2) Is there any between-year variation in the degree of predation within a plant population, and if so, which are the possible determinants? (3) Does the degree of predation differ among inflorescences due to size, among populations due to spatial arrangements or habitat differences, and among different parts of the distributional area of B. alpina? (4) How does seed predation affect seed-mediated gene flow of the plant? and (5) What are the selective pressures operating in the plant and its seed predators? Material and methods
Plant material and study site Bartsia alpina L. (Scrophulariaceae) is a common, perennial, hemi-parasitic, arctic-alpine herb growing in rich fens, alpine meadows, and creeping soil. Each clone usually produces 2-10 flowering ramets per year. The inflorescence is a raceme producing on the average 4.5 floral nodes (range 1-13) with paired flowers. The flowers are nectar-producing, zygomorphic, and pollinated by long-tongued bumble-
182 bees. Bartsia alpina has a mixed mating system (Molau et al. 1989); it is self-compatible but requires an external pollen vector for reproductive success. The fruit is a dehiscent, two-locular capsule containing on the average 48 winged seeds (Molau et al. 1989). Effective population size, as settled by maximum gene flow early and late in the season, exceeds 100 clones (genets) in the studied population. The field work was carried out during 1986-1988 in a large population of B. alpina in a subalpine open meadow of creeping soil at about 400 m a l t . near Abisko Research Station, northernmost Sweden. The population was about 130 m in square and had a density of 0.75 clones per square meter. Flowering started on June 20-25 and terminated on July 15-25 (range of 1986 1988). Primary seed dispersal occurred during the first half of September. The density of flowering genets was constant among years.
Sampling methods A total of 345 fruiting ramets were collected systematically (125 in 1986, 120 in 1987, and 100 in 1988) shortly after flowering ceased. Additionally, 111 ramets flowering in 1985 were collected in 1986. All capsules on a ramet were examined for seed predation in the following way: the capsules were opened and when larvae were still present, they were collected and conserved in 70% ethanol. The number of apparently undisturbed, non-predated seeds were counted. The position and number of holes in capsule walls and septa were noted. The actual node position (ANP) of attacked capsules on a ramet as well as of sterile, aborted flowers were noted. Additionally, in 1987 and 1988, based on experience from the first year, we distinguished between different larvae species found. For comparative purposes, the idea of relative node position (RNP) proposed by Best and Bierzychudek (1982) was adopted. All inflorescences were ascribed 5 nodes, which is the mean integer of nodes found in the population, and each actual node was assigned a position from 1 to 5 (see Molau et al. 1989). The length of the flowering season is almost the same for any individual B. alpina inflorescence, and large inflorescences produce more flowers per unit time than do smaller ones. Therefore, the number of capsules available to predators at a certain time is better reflected by relative than actual node positions. The severity of seed predation by host-specific predators has been recorded to vary strongly among habitats (Janzen 1975). To examine if this is the case also in B. alpina, ramets from a population at Mount Njulla at 900 m a l t . were collected in 1988. The numbers of (1) floral nodes per ramet, (2) completely unattacked ramets, and (3) attacked capsules, as well as the total number of capsules were counted and compared with the 1988 figures obtained from the population near Abisko Research Station at 400 m a l t . described above. To elucidate the question concerning the interaction of pollinators, predators, and plant attraction cues, samples of large infructescences (3 or more floral nodes) and small infructescences (two floral nodes) were systematically collected in the center of the population. The number of pollinated flowers, given as the number of seedsetting capsules, in relation to the total number of flowers produced was counted. Furthermore, completely unattacked ramets and the degree of predation of the remainder were recorded. It is known that relatively host-specific seed predators
may have alternative host plants and that these and the primary host are often congeneric (Janzen 1980), or otherwise closely related (Augspurger 1981). Bartsia alpina frequently grows in mixed stands with another Rhinanthoid species, viz., Pedicularis lapponica L. Therefore, it was natural to investigate if any relation between these two species occurred regarding their predators. From 100 systematically collected ramets of P. lapponiea predators and parasitoids were isolated, and the rate of attack by each was noted.
Predispersal retention of seeds Attacked capsules do not open to the same extent as unattacked due to damage of the capsule wall, and furthermore, frass from the predators often plug the exit. Seeds from attacked capsules enter the seed pool only when tamers fall to the ground. In order to determine when this takes place, ten infructescences produced in 1985 and ten produced in 1986 were marked with aluminum tags in August 1986 and their fate followed the subsequent years.
Germination experiments From randomly selected last-year infructescences 100 apparently undamaged seeds from unattacked and attacked capsules, respectively, were collected. Samples were taken in late August but on the same day to avoid errors caused by the rapidly declining germination capacity during the second summer (U. Molau, unpublished data). The seeds were sown on filter paper in Petri dishes and germinated under uniform conditions in a greenhouse.
Predators and parasites Imagos of the seed predators were caught in three different ways: (1) During periods of oviposition on Bartsia in 1987 and 1988, general samples of all Microlepidoptera fluttering in the study population were taken with a net. (2) Cages made of sticks and mosquito net were placed over entire clones of Bartsia in 1987 and 1988, prior to hatching of pupae buried in the soil. Clones with high levels of predation during the previous years (easily seen by the number of exit holes in the capsules of remaining infructescences) were sampled. The imagos appearing in the cages were collected and determined. (3) In August 1987, maturing inflorescences infested by seed predators were cut off and put together in plastic pots with sterilized soil. The larvae were observed when they left the capsules and dug into the soil to pupate. The pots were sealed with mosquito net and buried within a small Bartsia population near the Abisko Research Station, the top of the pots at ground level. After natural hibernation, hatched imagos were collected from the pots in June 1988. In order to collect imagos of predator parasites pupating in the capsules, a total of 100 infructescences from the previous season were gathered in June 1987. These were put in plastic vessels sealed with mosquito net, and after hatching the hymenopteran imagos were collected. Results
Natural history of predators and parasites In N Fennoscandia we found that seeds of Bartsia alpina are destroyed by two relatively host-specific predispersal
183
seed predators. The most frequent of the two is the microlepidopteran Aethes deutschiana Zett. (Lepidoptera-Tortricidae; det. by Mr. Ingvar Svensson, Kristianstad, Sweden). This species is common in the alpine parts of Fennoscandia (Brundin 1931; Benander 1950), but its larval stage and host plant were completely unknown until this study (I. Svensson, personal communication). Imagos of A. deutschiana were caught in all kinds of traps used in the experiments, and the species dominated in the general samples of fluttering Microlepidoptera in the Bartsia population. The dipteran seed predator, Gimnomera dorsata Zett. (Diptera-Scatophagidae; det. Dr. H. Andersson, Univ. of Lund, Sweden), was caught in much smaller numbers, probably due to its smaller size, allowing the imagos to escape from the traps. Also G. dorsata is abundant in the Abisko area, but its natural history was hitherto unknown (H. Andersson, personal communication). The closely related G. hirta Hendel is known as a seed predator in another subarctic-alpine rhinanthoid plant, viz., Pedicularis sceptrumcarolinum L. (Ryd6n 1933). In our field studies, both seed predators were also found in the fruits ofPedicularis lapponica. The visual results of attacks of the two kinds of predators look very much alike in Bartsia alpina. Oviposition normally starts when most of the plant population is about to start flowering, and is most intense during the first days. The violet lower bracts enclosing the young racemes are obviously acting as predator-attracting cues, since oviposition was never seen on purely vegetative shoots with leaves similar in shape but lacking the violet color. The eggs of both predators are placed in small clusters (usually 3-6 eggs) on the lower surface of the bracts (A. deutschiana eggs are whitish and ca. 0.3 mm diam., G. dorsata eggs are yellowish and somewhat smaller). In later stages of raceme development, eggs are placed also on pedicels and calyces, but with decreasing abundance and in smaller clusters. Oviposition ceases after about two weeks, normally around July 7. After a few days the eggs hatch and the larvae drill into the nearest developing ovary, leaving a minute entrance hole (0.1-0.2 m m diam.). Normally, there is but a single larva in each capsule, even though double (rarely triple) attacks are occasionally found. The young larvae drill into developing seeds and eat them from the inside out, but later, as the larvae grow bigger, seeds are eaten from the outside. The contents of a single capsule locule is sufficient for the development of one G. dorsata larvae, whereas A. deutschiana larvae each require the contents of an entire fruit. Thus, capsules attacked by A. deutschiana always have the septum between the locules broken through. If seed set in a capsule is insufficient, or if more than one A. deutschiana larva is predating on the same capsule, the larvae will leave the capsule, crawl upwards in the raceme, and enter the nearest unattacked capsule above. Such secondary A. deutschiana attacks are easily recognized by the large entrance holes near the base of the capsules. Gimnomera dorsata larvae are not capable of moving around in the raceme, and are thus not causing secondary attacks. At the same time, they are more efficient resource utilizers, leaving much less frass behind than does A. deutschiana. In the middle of August, the larvae leave the Bartsia capsules to pupate in the soil immediately beneath the host. Both predators gnaw a circular exit hole 1-2 mm diam. near the top of the capsule. Aethes deutschiana larvae crawl
Table 1. Seed predation in a Bartsia alpina population
No. of ramets examined No. of ramets attacked No. of capsules examined Attacked capsules (%) Lepidoptera attack (%) Diptera attack (%) Attacked capsules with parasitoids (%) Parasitoids on Lepidoptera (%) Parasitoids on Diptera (%)
1985
1986
1987
1988
111 90 774 38.2
125 101 919 44.3
120 104 936 48.0 35.0 13.0
100 96 659 83.3 45.4 38.0
10.9
12.4 83.8 16.2
100 8O
.~ 60 ~_ 40 20 0 1
2
3
4
5
RNP
Fig. 1. Predation level versus relative node position (RNP) over the years 1986 [B, 1987 [], 1988 []; - A - Mean downwards along the stem, whereas G. dorsata larvae drop to the ground directly from the exit holes. Both seed predators in Bartsia alpina are attacked by a parasitoid, the small parasitic wasp Seambus brevicornis Grav. (Hymenoptera-Parasitica-Ichneumonidae; det. Dr. Lars Hedstr6m, Univ. of Uppsala, Sweden) (Table 1). After consuming the contents of a predator larva, the parasite pupates in the capsule. The pupae hatch in June the following year, contemporary with the start of predator oviposition. Scambus brevicornis is polyphagous and common throughout Sweden, attacking a large number of primarily lepidopteran seed predators (L. Hedstr6m 1987, personal communication). Amplitude of predator attack Seed predation increased slightly from 38.2% to 48.0% attacked capsules during the first three years of study and then increased tremendously in the fourth year, where 83.3% of all capsules were attacked (Table 1). There was an increase from 1987 to 1988 in the larval predator fauna of 25% and 10.4% for G. dorsata and A. deutsehiana, respectively, suggesting that most of the increase in seed predation was due to larvae of G. dorsata. The seed predation was heaviest at the second node (RNP) in all years (Fig. 1), while in 1987 the three lowermost nodes suffered a proportionately higher predation than the uppermost two when compared to the other years (Fig. 1). In 1986 the number of remaining seeds in attacked capsules was 12.4_+0.7 (mean + S.E., N = 407). The percentage of seed predators killed by the parasitoid was almost the same in 1987 and 1988 (Table 1), suggesting that there is a linear relation between density of seed predator larvae and the parasitoid.
184 Variation in predation pressure The comparison between two B. alpina populations from different habitats and altitudes revealed that (1) the number of floral nodes per ramet did not differ significantly between Abisko, 400 malt., and Mt. Njulla, 900 m alt. (4.2, N = 100 vs. 4.1, N = 52), whereas (2) the rate of completely unattacked ramets (4.0%, N = 100 vs. 42.3%, N = 52), and (3) predation rate (83.3%, N = 6 5 9 vs. 20.8%, N=384) indicated a significantly lower predation level at Mr. Njulla (X 2 = 41.22, P < 0.001 and X 2 = 78.26, P < 0.001, respectively). Large inflorescences attracted both pollinators and predators to a greater extent than small ones. The fraction of pollinated flowers was significantly higher in large inflorescences (X z = 13.3, P
