Short-term indirect interactions between two moth ( Lepidoptera : Noctuidae ) species mediated by shared parasitoids : The benefit ofbeing scarce

Despite the impact of parasitoids on insect populations being extensively studied, indirect parasitoid-mediated effects remain rarely documented in natural communities. We examined the influence of shared parasitoids on the interactions between two functionally monophagous moths, Nonagria typhae and Archanara sparganii. The moths showed a considerable variation in terms of relative abundance and the degree of phenological synchrony between the species. On average, parasitism levels caused by shared parasitoids did not differ between the two host species. Relative parasitism levels of the two hosts, however, varied considerably among different samples. Percentage parasitism of the scarcer species, A. sparganii, thus could not be fully explained by that of the dominant species, N. typhae. The results indicated that A. sparganii may benefit from the presence of N. typhae. In particular, both low relative density as well as high phenological synchrony with N. typhae reduced parasitism levels in A. sparganii. The case thus indicates the presence of parasitoid-mediated indirect effects between the coexisting herbivores. The patterns of host use observed in this study are consistent with the scenario of frequency-dependent host use caused by changes in parasitoid behavior. Such a host use by parasitoids is suggested to promote numerical stability and coexistence of the moth species in the system studied.


INTRODUCTION
An interaction between two species is called indirect when the effects of one species on another are mediated by some third species (Strauss, 1991;Wootton, 1994).Natural-enemy-mediated indirect effects may arise either through functional (short-term effects) or numerical response (long-term, or transgenerational effects) of the enemy to the prey population.Enemy-mediated indirect effects have been suggested to play a significant role in natural communities affecting community structure and population dynamics of the species involved (Jeffries & Lawton, 1984;Holt & Lawton, 1994;Müller & Godfray, 1999).However, the empirical evidence on such effects is still scarce (Bonsall & Hassell, 1999;Chaneton & Bonsall, 2000).In a recent review, Chaneton & Bonsall (2000) found as few as 21 studies addressing questions about, or interpreting results, in the context of apparent competition.The evidence on horizontal, positive indirect effects in communities is even more scattered and comes mainly from herbivore-mediated interactions between plants (Houston et al., 1993;Olofsson et al., 1999).
Interspecific interactions between coexisting herbivo rous insects can be mediated both by lower (host plant) and higher (natural enemies) trophic levels.While hostplant-mediated effects are usually mutually negative (resource competition), the type and strength of indirect interactions mediated by shared predators or parasitoids are less obvious.The outcome -are the interactions mutu ally positive (apparent mutualism), mutually negative (apparent competition), or non-reciprocal (indirect amensalism) -is suggested to depend on various factors, e.g.foraging behavior of the natural enemies, relative feeding habits or competitive abilities of victim species (Jeffries & Lawton, 1984;Holt, 1987;Holt & Kotler, 1987;Holt & Lawton, 1994;Abrams et al., 1998).
In the case of herbivorous insects, parasitoids have probably the greatest potential of mediating indirect effects between coexisting species.The type and strength of parasitoid-mediated indirect interactions are pre sumably sensitive to the composition of parasitoid com munity.This is because of a high diversity of lifehistories (e.g.koinobionts vs idiobionts, ectoparasitoids vs endoparasitoids) and diverse behavioral repertoire (host preference, host switching etc.) among parasitoids.Empirical studies on various systems are therefore required to understand how widespread parasitoidmediated indirect effects may be in natural communities, and which factors influence the expression of such effects.While manipulative studies are necessary to reveal causal relationships, an analysis of correlative data would estimate the range and relevance of the effects in the field.
In the present field study, we demonstrate a pattern consistent with indirect, parasitoid-mediated interactions between a pair of insect herbivores exploiting a common host plant.The species studied are functionally mono phagous and their parasitoids lack alternative host species in the study areas.In its relative isolation, the system is thus well suited for examining indirect interactions in the field.We document variation in relative parasitism levels of the two host species, and analyze possible factors responsible for the patterns observed.Attention is paid to the possible role of indirect effects mediated by variations in relative abundance and relative phenological distribu tion of the herbivores.Finally, we discuss possible conse quences of the detected indirect interactions on the population dynamics of the moth species and community structure.

Study species
The study system was based on Typha latifolia L. (Typhaceae), a rhizomatous perennial plant up to 3 m in height forming dense stands in damp sites.Larvae of four moth species have been found feeding on leaves and stems of T. latifolia (Teder et al., 1999;Teder & Tammaru, 2002).This study was, however, restricted to the two most abundant species, Nonagria typhae Thunberg and Archanara sparganii Esper (Lepidoptera: Noctuidae).The proportion of other species remained < 1% of all sampled larvae.The species studied are close taxonomically and have a high resemblance in feeding biology and phenology (pers.obs.).They overwinter as eggs that hatch in spring.At the beginning of the season larvae feed on the aerial roots of the host plant, while later they switch to endophytic feeding in the shoots (Galichet et al., 1992).One larva usually feeds on more than one shoot during its development.Larvae of N. typhae are monophagous on T. latifolia, while larvae of A. sparganii may also use some other species (Skou, 1991), however, in the studied areas both species were apparently functionally monophagous on T. latifolia.At the end of July or at the beginning of August larvae pupate (A.sparganii somewhat earlier than N. typhae) inside the shoot or between the leaves of the host plant.The pupal period lasts for about one month in both species.
Two principal species of solitary parasitoids, the koinobiont Spilichneumon limnophilus Thomson and the idiobiont Chasmias paludator Desvignes (Hymenoptera: Ichneumonidae), were found to parasitise larvae and pupae of both N. typhae and A. sparganii.A third species parasitising both moths, Vulgichneumon saturatorius L., was rare in most years.These three species were treated as the shared parasitoids in the analy ses.The most numerous parasitoid of N. typhae, Exephanes occupator Gravenhorst, parasitises A. sparganii only occasion ally (Teder et al., 1999): in most years no individual A. spar ganii was parasitised by this species.Other parasitoids treated here as non-shared, accounted for < 1% of the total parasitism.Like their hosts, studied parasitoids have univoltine life cycles.Adult females overwinter and lay their eggs on moth larvae in late spring / early summer [E.occupator (Hinz & Horstmann, 2000;pers. obs.) and S. limnophilus (pers.obs.)] or on fresh pupae in July and early August [Ch.paludator (Hinz, 1983;Hinz & Horstmann, 1999)].Adult wasps emerge more or less simultaneously with unparasitised moths (E.occupator some what earlier).All three parasitoids are oligophagous and are also known to parasitise some other moths (Rasnitsyn & Siitan, 1981;Hinz & Horstmann, 1999, 2000) which, however, appar ently constitute only a minor fraction of hosts in the habitats studied.

Study areas
This work was conducted in 7 consecutive years (1995)(1996)(1997)(1998)(1999)(2000)(2001)) in southeastern Estonia.Three habitats differing with respect to Typha distribution pattern were examined.The habitat studied in 1995-1996 was characterized by a dense, almost monospe cific stand of T. latifolia.The vegetation of the habitat examined in 1997-1998 was more heterogeneous with patches of T. lati folia and Carex elata Bell.ex All.alternating.Both these habi tats, with the areas of 3 and 4 ha, respectively, were located in the town of Tartu (58°22' N, 26°45' E), on the flooded meadow of the Emajogi River.The third habitat, studied in 1998-2000 was the most heterogeneous where patches of T. latifolia (mostly from 0.01 to 0.05 ha, at small ponds, ditches, or other damp places) were separated by distances of 0.2-1.0km.This study area was located in an agricultural landscape, close to the Lake Pangodi (58°12' N, 26°35' E), 20 km southwest of Tartu.The samples collected from Tartu and Pangodi in 1998 were treated as independent in the analyses.

Sampling
Field populations were sampled to obtain data on the distribu tion of moths and the levels of parasitism.Sampling was con ducted at the end of July and/or the beginning of August, when moths were in the pupal stage.A variable fraction of A. spar ganii pupae (2-51%) had eclosed by the day of collection.Simi larly, a variable, but relatively small fraction of N. typhae specimens (4-21%) were collected as larvae.The fractions of individuals collected before or after the pupal period in a par ticular sample were regarded as indices reflecting phenology of the species (see below for details).Each year, subsamples from 12-33 plots (173 in total) were collected.In the habitats located in Tartu, subsamples were taken from plots of 2 x 2 m, 3x3m or 4 x 4 m (depending on larval abundance in the particular year), while in the most heterogeneous habitat (Pangodi) at least 30 individuals (N.typhae + A. sparganii) were sampled per each patch.From the study plots, moth pupae were collected by inspecting all shoots carefully.The pupae of both species were relatively large (2.0-3.5 cm in length), potential pupation sites were limited, and damaged shoots were easily distinguishable from undamaged shoots.Therefore, it was possible to collect nearly all pupae from the plots.The number of pupae collected in one sample varied from 200 to 2294 (5736 individuals in total; 757 A. sparganii and 4979 N. typhae).Pupae were stored in Petri dishes until eclosion of adult moths or parasitoids.For A. sparganii individuals collected as exuviae, it was always pos sible to determine whether a moth or a parasitoid, and in most cases, which of the parasitoids, had emerged.The fate (parasitised / unparasitised) of the specimens that died in the laboratory in the early pupal stage (mainly N. typhae) could not be determined; they were thus omitted from calculations and analyses (5-10% of specimens each sample).

Data analysis
Levels of parasitism in A. sparganii and N. typhae were ana lysed with respect to relative abundance of the moths and their relative phenology.As the purpose of this study was to examine parasitism-mediated indirect effects, all the calculations and analyses were restricted to plots in which both A. sparganii and N. typhae were found.Data of such plots were pooled within samples for subsequent analyses.For the same reason, only parasitism caused by shared parasitoids was considered in the analyses.In A. sparganii, shared parasitoids accounted for most of the parasitism.In contrast, for N. typhae a considerable frac tion of individuals were parasitised by the non-shared E. occu pator.Although multiple parasitism of hosts previously parasitised by E. occupator was unlikely in this system, these individuals were re-classified as non-parasitised for the analyses.Such an approach was taken assuming that the shared parasitoids, S. limnophilus and Ch.paludator, presumably make their decisions on relative abundance of the two host species on the basis of overall host density (i.e.unparasitised + already parasitised hosts) rather than density of unparasitised hosts.To derive an index describing differences in phenology of the two moths, the study samples were ranked according to 1) the percentage of adults among A. sparganii (found as exuviae) on the day of sampling, and 2) the percentage of larvae among N. typhae on the day of sampling.The value "1" was assigned to the sample of the highest proportion of adults among A. sparganii, and to the sample of the lowest proportion of larvae among N. typhae, both these ranks indicating the earliest pheno logies.Accordingly, the value "8" indicated the latest phenolo gies.The relative phenology index for a particular sample was calculated by subtracting the rank of A. sparganii from the cor responding rank of N. typhae.The larger was the index obtained in this way, the larger was the difference in phenological distri butions of the two moths.
The rationale of this relative phenology index was in describing the degree of temporal overlap, or co-occurrence of the vulnerable stages of the two host species.This may be rele vant in the context of indirect interactions.As predicted by the optimal foraging theory, higher abundance of the preferred host should increase parasitoid's selectivity.Accordingly, when the peaks of vulnerable A. sparganii and N. typhae overlap, more larvae of the preferred host, N. typhae, are available, and thus, a larger proportion of larvae of the less preferred species, A. sparganii, should escape parasitism.In the opposite case, when the phenological distributions of A. sparganii and N. typhae differ, a larger proportion of vulnerable A. sparganii would fall into the period of lesser choosiness of the parasitoids.
Logistic regression analysis (PROC GENMOD; SAS Institute Inc., 1995) was applied to examine whether the distribution of parasitism between A. sparganii and N. typhae differs in the study samples.Binomial probability distribution was assumed, Fig. 2. The relationship between parasitism levels of A. sparganii and N. typhae.Only parasitism by S. limnophilus is con sidered; x -and y -error bars represent 95% confidence intervals for the binomial parameter (parasitised/nonparasitised).logit was chosen as the link function, DSCALE option was applied to correct for overdispersion.Incidence of parasitism (parasitised/nonparasitised) was used as the response variable, and "sample", "host species" and "sample x host species" were used as the independent effects.Linear regression was used to examine the effects of relative abundance of the moths and their relative phenological distribution (see above) on their relative distribution of parasitism at the level of sample means.The dependent variable, relative parasitism of the two host species was expressed as the ratio of the fractions of parasitised A. sparganii and N. typhae per sample.For example, if in a particular sample, percentage parasitism of A. sparganii was 12% and that of N. typhae was 30%, the corresponding index of relative para sitism was 12 / 30 = 0.4.The use of this relative measure was chosen to eliminate the effect of overall parasitoid abundance on the parasitism levels of A. sparganii.Analogously, relative abundance of the two moths was calculated as the ratio of the numbers of A. sparganii and N. typhae collected in the same sample.

RESULTS
Among the samples studied, parasitism levels were highly variable in both of the moth species.Mean values of total percentage parasitism fluctuated from 5.8% to 38.5% in A. sparganii and from 23.5% to 68.1% in N. typhae (Fig. 1).When only shared parasitoids were considered, percentage parasitism of N. typhae dropped to the range of 7.0% to 33.6%, whereas parasitism of A. sparganii, being caused mainly by shared parasitoids, changed only negligibly (Fig. 1).Variation in the para sitism levels by individual parasitoid species was also considerable (e.g.parasitism by S. limnophilus fluctuated from 2.6% to 19.2% in N. typhae and from 1.4% to 22.1% in A. sparganii).
Parasitism levels of the scarcer species, A. sparganii, tended to covary with those of the dominant species, N. typhae, the correlation presumably being explained by the overall abundance of the parasitoids.A positive and significant association was observed when parasitism by S. limnophilus was analysed (Fig. 2).For parasitism by Ch. paludator no association was observed between the parasitism levels of A. sparganii and N. typhae.The overall levels of parasitism by the shared parasitoids did not differ between the two moth species (Table 1).In contrast to this "overall equality", parasitism levels were not equal in particular study samples.For example, in 1995 parasitism percentage of N. typhae by shared parasitoids exceeded that of A. sparganii more than 3 times whereas in 1998 (Tartu -site) A. sparganii appeared to suffer from parasitism levels twice higher than N. typhae (Fig. 1).The among-sample differences in the relative parasitism of the two host species were statis tically confirmed by a highly significant "sample x host species" interaction (Table 1).
One possible factor able to explain these remarkable differences in relative parasitism levels is the variable relative abundance of the host species.Though A. sparganii was less abundant in all samples, relative abundance of A. sparganii and N. typhae considerably fluctuated among samples.The index of relative abundance of the two species (the number of A. sparganii divided by the number of N. typhae) fluctuated from 0.03 (1996) to 0.49 (2001).Linear regression of mean values indicated that relative abundance of the moths affected their relative parasitism levels.The effect of relative abundance on relative parasitism levels was positive and significant when parasitism by S. limnophilus was analysed (Fig. 3).Similarly, a positive though not significant trend was observed also when parasitism by Ch. paludator (R2 = 0.19, NS) was analysed.Another factor that may have differential influence on the availability, and thus, parasitism levels of the two moths, is their relative phenological distribution.Corre spondingly, the relative phenology index (see Methods) correlated positively with relative parasitism levels caused by S. limnophilus (Fig. 4).Again, the relationship between relative phenology index and relative parasitism levels by Ch. paludator (R2 = 0.11, NS) was also positive but non-significant.Mean values of relative abundance and relative phenology index appeared to be correlated (R = 0.81).Unfortunately, a two-way ANOVA examining the effects of relative abundance and relative phenology jointly would not have been statistically meaningful due to small sample size (8 samples).

DISCUSSION
The present results revealed a pattern that is consistent with the presence of indirect, parasitoid-mediated interac tions between two herbivores.We showed that relative parasitism levels of the two coexisting host species, A. sparganii and N. typhae, considerably varied among dif ferent samples.The present analyses revealed two factors that may contribute to the patterns observed via func tional responses of the parasitoids.In particular, relative parasitism levels of the two moths correlated both with relative abundance of the two moths as well as the differ ence in their phenologies.Although we cannot strictly prove which of the two factors is more important in deter mination of parasitism levels of A. sparganii, the conclu sions would be similar: A. sparganii appeared to benefit from the presence of N. typhae, or in other words, escape parasitism by its relative scarcity.The case is one of very few (see also Bonsall & Hassell, 1997;Müller & Godfray, 1997) providing evidence for indirect interac tions in parasitoid-involving systems at a short time-scale (see Holt & Lawton, 1994).Unfortunately, however, the assymetric distribution of the abundances does not allow us to judge about the symmetry of this interaction (Chaneton & Bonsall, 2000).
The effect of relative abundance is likely to be explained by the consequences of frequency-dependent host use by the parasitoids.Such a pattern of host use may result either from the host-frequency-dependent changes in a) host's anti-parasitoid behaviour, or b) for aging behaviour of the parasitoids (i.e. a change in host preference) (for a review, Sherratt & Harvey, 1993;van Alphen & Jervis, 1996).The mechanisms of the first type are unlikely in the present system: given the solitary, con cealed life-style of the herbivores studied, an increase in the density hardly has a chance to change their defensive behaviour.A change in host preference in parasitoids is a more likely explanation.In the system studied such a switching (sensu Murdoch, 1969) might arise as a conse quence of changes in host composition: the higher the A. sparganii density, the greater the possibility for parasi toids to learn stimuli deriving from this species, although alternative scenarios are possible (Holt, 1983;Sherratt & Harvey, 1993).
From the parasitoids' point of view, the relative abun dance of a pair of hosts is not necessarily determined by their absolute numbers: host individuals of different ages are rarely equally vulnerable to parasitoid attacks (Briggs & Latto, 1996;Benrey & Denno, 1997).The risk of para sitism for the coexisting hosts might thus also depend on the degree of synchrony in hosts' windows of vulnerabil ity.If parasitoids forage for the hosts in a frequencydependent manner, the rarer host may reduce parasitism risk by adjusting its vulnerable stage to occur simultane ously with the more abundant host.The present results are consistent with this scenario: smaller differences in the phenological distributions of the two hosts implied a reduced parasitism risk for A. sparganii.Such a pattern may be seen as another indication of a change in host preference in the system studied.
The present system appears to meet the necessary pre conditions for switching behaviour to evolve: 1) relative abundance of different prey types should vary either spa tially or temporally and 2) predators should be mobile (Cornell, 1976).In the present system, the relative abun dance of the two moths varied among samples; shared parasitoids are extremely mobile and can be met even in very isolated patches (Teder & Tammaru, pers. obs.).Moreover, both moths are quite abundant in absolute terms, they seem to be well suitable for the parasitoids, and their feeding biologies and phenologies are similar.Parasitoids are thus faced with an actual choice of dif ferent host species, and switching may appear to be favoured in this system.Such a complex of preconditions is seemingly rarely fulfilled in most natural systems as reflected by rare documentation of changes in host prefer ence in the field (see, however, Kato, 1994;Pike et al., 1999), whereas the phenomenon is repeatedly demon strated under laboratory conditions (Cornell & Pimentel, 1978;Chow & Mackauer, 1991;Drost & Cardé, 1992).
Frequency-dependent host use has been suggested to have significant consequences for the structure and dynamics of multispecies communities by promoting sta bility and coexistence of the species (Holt & Lawton, 1993;Bonsall & Hassell, 1999;Hassell, 2000).One of the most obvious consequences of parasitism in the studied system is its potential to stabilize population dynamics of the scarcer moth, A. sparganii.At lower den sities, parasitism levels of A. sparganii appeared to be lower which should allow for A. sparganii populations to increase.As N. typhae populations maintain the density of parasitoids continuously high (N.typhae was the more abundant host species in all samples), the response of parasitoids to an increase in A. sparganii density occurs without a delay, a mechanism widely appreciated as pro moting stable population dynamics in the host (Holt & Lawton, 1993).In the studied system, positively densitydependent parasitism accompanying with a change in host preference allows A. sparganii to increase when rare thus facilitating coexistence of the host species (see also Holt & Kotler, 1987).Mechanisms promoting stability for the populations of N. typhae remain unclear.

Fig. 1 .
Fig. 1.The levels of shared and total parasitism in A. sparganii and N. typhae in the study samples (1998m -sample col lected in Tartu, 1998h -sample collected in Pangodi, see text for details).Sample sizes are presented above the bars.

Fig. 3 .
Fig. 3.The relationship between relative abundance of the moths (the ratio of the numbers of A. sparganii and N. typhae collected in the same sample) and relative parasitism (the ratio of fractions of parasitised A. sparganii and N. typhae per sample) at the level of means.Only parasitism by S. limnophilus is considered.

Fig. 4 .
Fig. 4. The relationship between relative phenology index (the larger the index, the larger the difference in phenological distributions of the two species) and relative parasitism (the ratio of the fractions of parasitised A. sparganii and N. typhae per sample) at the level of means.Only parasitism by S. limnophilus is considered.

Table 1 .
The results of logistic regressions (significance tested by type I analyses) examining the differences in the dis tribution of parasitism between A. sparganii and N. typhae.