Infection and parasitism rates
We collected a total of 4214 and 632 live larvae and larval cadavers, respectively, across all sites over the 3 years, from which we detected no pathogens, LdNPV only, E. maimaiga only and both pathogens at 22, 0, 7 and 8 sites, respectively. We detected parasitoids at 16 of the 37 sites. The frequency and prevalence of both pathogens and parasitoids across 2005–2007 are presented in Table 1. Across years, E. maimaiga was consistently more abundant than other natural enemies with infection ranging from 16·0% to 45·5% at sites where it was present. In contrast, LdNPV prevalence was much lower at 0·8–2·9%. Parasitoids reared were predominantly the generalist-introduced tachinid fly (>90%) C. concinnata (Table S2), and parasitism ranged from 1·6% to 11·9%.
Table 1. Frequency and prevalence of pathogens and parasitoids attacking Lymantria dispar from 2005 to 2007a
|Year||No. of sites||No. of sites L. dispar larvae and cadavers collected (range of no. larvae plus cadavers collected/site)||Mean number of larvae and cadavers collected per site (±SE)||Entomophaga maimaiga||L. dispar nucleopoly-hedrovirus||Parasitoids|
|No. of sites present||% Infection (mean ± SE)||No. of sites present||% Infection (mean ± SE)||No. of sites present||% Parasitism (mean ± SE)|
|2005||9||4 (3–162)||22·0 ± 17·8||2||45·5 ± 8·8||2||2·9 ± 0·4||1||1·9 ± –|
|2006||12||12 (1–320)||111·1 ± 34·3||6||18·6 ± 8·3||5||2·8 ± 0·9||7||11·9 ± 2·6|
|2007||31||28 (1–348)||106·9 ± 17·0||14||16·0 ± 5·6||5||0·8 ± 0·2||11||1·6 ± 0·5|
The presence of infection at our field sites was not significantly related to the distance from prior releases of pathogens or subsequent recoveries (that were predominantly near releases) (E. maimaiga: G2 = 0·07, d.f. = 1, P = 0·70; LdNPV: G2 = 0·43, d.f. = 1, P = 0·53). We also did not observe a significant effect of the time between releases or recoveries and the year of our study (E. maimaiga: G2 = 1·76, d.f. = 1, P = 0·18; LdNPV: G2 < 0·01, d.f. = 1, P = 0·98) or a significant distance-by-time interaction (E. maimaiga: G2 = 0·04, d.f. = 1, P = 0·84; LdNPV: G2 = 0·46, d.f. = 1, P = 0·50), suggesting that these biological control releases did not account for the observed patterns of pathogen distribution (Fig. 2). Although some of our sites were within 0·5 and 6·1 km of prior releases of E. maimaiga and LdNPV, respectively, the median distances from sites where pathogens were detected to release sites were much greater (65·2 and 25·6 km, respectively).
Figure 2. Predicted model probabilities (solid black line, with 95% confidence intervals as dashed black lines) of the presence of Entomophaga maimaiga (a) and Lymantria dispar nucleopolyhedrovirus (LdNPV) (b) relative to the distance from known release or recovery sites (see Fig. S2). The histograms represent the number of sites with E. maimaiga (a) or LdNPV (b) present (black bars) or absent (gray bars).
Download figure to PowerPoint
The background male moth abundance (a proxy for larval density) in the prior year (G2 = 7·9, d.f. = 1, P < 0·01) and in the year of study (G2 = 8·5, d.f. = 1, P < 0·01) was a significant predictor of infection by E. maimaiga. However, only the background male moth abundance in the year of study was a significant predictor of infection by LdNPV (G2 = 7·7, d.f. = 1, P < 0·01). Infection by E. maimaiga was more likely than not to occur at male moth abundances of >81 and >143 from the prior and current year, respectively, while infection by LdNPV was more likely than not to occur at male moth abundances of >164 from the year of study. Thus, only for E. maimaiga did the predicted probability of infection at a site increase with increasing male moth abundance from the prior year. There was no association between infection by LdNPV and the subsequent change in male moth abundance from the prior year to the survivors from the year of study (G2 = 0·9, d.f. = 1, P = 0·33). However, there was a significant association between infection by E. maimaiga and subsequent decrease in male moth abundance (G2 = 68·9, d.f. = 1, P < 0·01). The odds ratio indicated that at sites where E. maimaiga infection was present, the adult male population was 2·1 (95% CI = 1·6–2·7) times more likely to decrease from the prior year (i.e. male parent population of the larvae we sampled) to the year of study (i.e. the adult male populations that subsequently survived from the larval populations that we sampled).
When considering the cumulative prior history of male L. dispar abundance at each site, there was a significant relationship between the number of years in which male moths exceeded 100 per trap and infection by both E. maimaiga (G2 = 11·9, d.f. = 1, P < 0·01) and LdNPV (G2 = 4·0, d.f. = 1, P = 0·04), while the number of years for which male moths exceeded 10 per trap was a significant predictor for only E. maimaiga (G2 = 5·7, d.f. = 1, P = 0·02) (Fig. 3). The respective logistic regression models for E. maimaiga and LdNPV differed. The predicted probability of E. maimaiga infection is c. 0·8 when the site has exceeded the 100-moth threshold for only one prior year, whereas for LdNPV, the same probability is predicted to occur when the site has exceeded the 100-moth threshold for four prior years. A similar probability (0·8) of E. maimaiga infection was also predicted when a site has exceeded the 10-moth threshold for c. 5 years (Fig. 3). For both pathogens, exceeding the 1-moth threshold was a non-significant predictor of infection (P > 0·4 for both).
Figure 3. Predicted probability of infection or parasitism based on prior Lymantria dispar population density history for Entomophaga maimaiga (a), L. dispar nucleopolyhedrovirus (LdNPV) (b) and larval parasitoids (c). The dotted, dashed and solid black lines represent the 1-, 10- and 100-moth population thresholds, respectively. Lines with an asterisk denote significant relationships.
Download figure to PowerPoint
Quantifying the rate of spread in an invading species can be challenging because of the difficulty in obtaining the spatial and temporal data that are required to estimate invasion speed. In the case of E. maimaiga, which was shown to be significantly related to the cumulative prior history of L. dispar at two population thresholds (the 10- and 100-moths-per-trap thresholds; Fig. 3), it is possible to relate the change in infection levels based on the prior host history using L. dispar spread rates and rate of population transition time at the time of our study, which are feasible to estimate (Tobin & Whitmire 2005; Tobin, Liebhold & Roberts 2007a). For example, c. 50% of larvae from sites were infected with E. maimaiga when the prior L. dispar population exceeded 100 moths per trap for <1 year, and 50% of larvae from sites were infected with E. maimaiga when the prior L. dispar population exceeded 10 moths per trap for c. 3 years (Fig. 3). At the time and place of our study, L. dispar was spreading at a rate of c. 12·2 km year−1, while the average distance between the 10- and 100-moth population thresholds was c. 37·1 km (Fig. S3); thus, it took c. 3 years (37·1 km/12·2 km year−1) for L. dispar populations to transition from a 10-moth to a 100-moth population threshold at our field sites. Because there was a similar 3-year lag in E. maimaiga infection (Fig. 3) when considering these two L. dispar thresholds, it is possible that E. maimaiga moved at the same speed as L. dispar at our study sites but was lagged in space.
For the 12 sites that were sampled in successive years, owing to the variability in the densities of host colonization when sampling and the overall low densities, no general trends were evident. However, in a few instances, we caught populations as infection prevalence increased, e.g. in 2006 at Rocky Arbor, we found only two larvae infected with E. maimaiga on the last of eight sample dates. (A total of 265 larvae were collected in 2006.) The next year, E. maimaiga infections began at 9·6% on 3 June and ranged from 33·7% to 66·7% on the three successive sampling dates.
Using stepwise logistic regression, three of nine climate variables were significantly associated with infection by E. maimaiga (Fig. 4): total April rainfall (positively associated; G2 = 23·9, d.f. = 1, P < 0·01), May temperature (negatively associated; G2 = 288·2, d.f. = 1, P < 0·01) and June temperature (positively associated; G2 = 171·6, d.f. = 1, P < 0·01). For infection by LdNPV, only April temperatures (G2 = 18·4, d.f. = 1, P < 0·01) and June temperatures (G2 = 7·3, d.f. = 1, P < 0·01) were significantly positively associated with rates of infection (Fig. 4).
Figure 4. The association between April rainfall, May temperature and June temperature and the proportion of Entomophaga maimaiga- or Lymantria dispar nucleopolyhedrovirus (LdNPV)-infected larvae per site (top two panels). Lines represent fitted logistic regression curves and 95% CI. Only significant associations are shown. The daily mean temperature and precipitation across all sites and years, April to June, are shown in the bottom panel with the predicted periods of 5% egg hatch and 95% completion to second instar (Régnière & Sharov 1998).
Download figure to PowerPoint
Rates of larval parasitism by all tachinids were not associated with male moth abundance from the prior year (G2 = 0·1, d.f. = 1, P = 0·77), nor was there a significant association between parasitism rate and the change in male moth abundance from the prior year to the year of the study (G2 = 0·6, d.f. = 1, P = 0·42). When considering the cumulative prior history of male L. dispar abundance at each site (Fig. 3), there was a significant negative relationship between the number of years for which male moths exceeded 1 (G2 = 18·4, d.f. = 1, P < 0·01), 10 (G2 = 5·4, d.f. = 1, P = 0·02) and 100 (G2 = 51·7, d.f. = 1, P < 0·01) moths per trap and parasitism. Therefore, rates of larval parasitism were highest when L. dispar history was the shortest but declined rapidly with longer histories of L. dispar presence.
Interactions among pathogens and parasitoids
Parasitism and infection proportions were frequently relatively low, as would be consistent with both recently invading hosts and natural enemies (Table 1). In our attempts to investigate the extent that different natural enemy species could successfully co-attack the same larval hosts, we did not observe any instances where both fungal and viral pathogens reproduced within the same host. We collected five L. dispar among the 762 infected by E. maimaiga, from which both fungal conidia were produced and C. concinnata successfully developed, and one larva, of the 26 infected by LdNPV in which C. concinnata also successfully developed.
When investigating the relationship between pathogen infection and parasitism, we observed a natural break in the data when considering sites with <30% of pathogen infection (1369 larvae) and ≥30% (1088 larvae) and thus considered these two group separately in our analysis. In both groups, there was a significant difference between proportion of infection and parasitism (G2 = 36·7, d.f. = 1, P < 0·01 in the >30% group; G2 = 318·3, d.f. = 1, P < 0·01 in the ≥30% group). However, the relative differences for these two groups varied; at sites with <30% infection, larvae were only 2·0 (95% CI = 1·6–2·5) times more likely to be infected relative to being parasitized, while at sites with ≥30% infection, larvae were 34·1 (95% CI = 23·1–50·2) times more likely to be infected relative to being parasitized (Fig. 5).
Figure 5. Relationship between pathogen (Entomophaga maimaiga plus LdNPV) and parasitoid levels at sites with combined pathogen infection rates of <30% or ≥30%. Although there were significant differences between pathogen infection and parasitism (as denoted by asterisks) in both infection level groups, the difference was greatest when the rate of infection was ≥30%.
Download figure to PowerPoint