We collected S. noctilio in New York, Pennsylvania and Vermont in 2013, 2014 and 2015. We searched for stands of hard pines (P. resinosa and P. sylvestris) during spring and summer, noting the locations of trees with emergence holes and resin drippings indicating S. noctilio attack in the previous season. Trees attacked by S. noctilio were difficult to find, with > 80% of stands showing no signs of woodwasp presence. In total, we located active S. noctilio populations in six locations in New York (MF: Montour Falls, Schuyler County, 42.3354°N, 76.8138°W), Pennsylvania (DE: Delmar, Tioga County, 41.7209°N, 77.3772°W; MI: Middlebury, Tioga County, 41.8416°N, 77.4072°W, BL: State Game Lands 276 in Blacklick, Indiana County, 40.4886°N, 79.1070°W; CL: Clarion County, 41.1777°N, 79.2269°W) and Vermont (UN: Underhill, Chittenden County, 44.4847°N, 72.9656°W) across three years. As is common in the region, most sites primarily contained one tree species: three stands contained P. sylvestris only (MF, DE, MI), two stands contained P. resinosa only (BL, CL) and one stand contained both (UN), though P. resinosa was more abundant (Table 2). We recognize that species and site are necessarily confounded in this study due to the rarity of naturally-attacked hard pine stands on the landscape. We still chose to include comparisons of tree species in our analysis because each species occurred at multiple sites dispersed throughout the study area.

During late June or early July 2014–2016, we visited previously-identified sites to cut trees attacked in the prior season. All trees were dead or dying at the time of cutting. Over three years, we felled 38 attacked trees, recording GPS locations and diameter at breast height (dbh) for each. After felling, we cut each stem into ~ 1 m-long bolts, discarding the top of the tree (diameter < ~ 2 cm). Bolts were labeled individually to record their position relative to the ground and transported to emergence containers at Dartmouth College, Hanover, NH.

Wasp-infested logs were placed in 55-gallon laminated cardboard emergence drums and stored indoors (ambient laboratory temperature ~ 21 °C). Drum openings were covered with fine mesh to prevent insect escape. Drums were checked every 1–3 days during peak emergence (mid-July to September) and at least twice a week thereafter until several weeks had passed with no new emergences (late October or early November). At the end of the first emergence season, we removed dead insects, then stored bolts in the laboratory until the following May, at which point we resumed regular checks. All emerging insects were collected, including S. noctilio, the native S. nigricornis and hymenopteran parasitoids. Due to low overall numbers, rhyssine parasitoids (R. persuasoria and R. lineolata) and kleptoparasitoids Pseudorhyssa spp. were combined for analysis and referred to as “rhyssines”. For S. noctilio and S. nigricornis, we measured body length (excluding the ovipositor), then dissected each individual to check for nematodes. Nematodes from a subset of wasps were cultured and confirmed as D. siricidicola when they were sequenced as part of a study by Fitza et al. (2019). For 176 of the females, we counted the total number of eggs and, for 36 haphazardly selected individuals, we measured the length of five eggs per female using a compound microscope.

Collected bolts were weighed and measured for length and diameter at each end. For each bolt, we calculated the surface area and volume from length and diameter. Before we placed them in emergence drums, we measured wood moisture content at five locations along each bolt (~ 0.8 cm depth) using a Delmhorst RDM-3 moisture meter (Delmhorst Instrument Co., Towaco NJ) and averaged these measurements for analysis. Occasional measurements above or below the operating range of the device (6–60%) were recorded as 6% and 60%, respectively.

After insects finished emerging in the second year (at least two months with no further emergence), we dissected a subset of bolts (2–3 per tree). We counted resin drippings and emergence holes, then removed the bark and cambium to count oviposition sites (attacks) and the number of holes (drills) per attack using an illuminated tabletop magnifier (5×) and hand lens (10×–20×) as needed, following methods established by Lombardero et al. (2016). For each attack, we noted the presence or absence of lesions (Suppl. material 1: Figs S1, S2), which are thought to be indicative of a polyphenol defense response by the tree (Coutts and Dolezal 1966; Lombardero et al. 2016). Since lesions were assessed two years after bolts were collected and may have faded over time, we considered this to be a conservative estimate of lesion formation.

For each bolt, we estimated the number of eggs laid, based on the number of attacks and drills per attack following Madden (1974) and Haavik et al. (2015):

# eggs = 0.01∙single drills + 0.68∙double drills + 1.55∙triple drills + 2.22∙quadruple drills

After dissection, we cut each bolt into three equal lengths (avoiding knots) to expose fresh surfaces. We estimated the percentage of the cross-sectional area colonized by bluestain (ophiostomatoid) fungi by outlining visible bluestain on the cut surface of each bolt and photographing the surface (see Suppl. material 1: Fig. S2). Area quantification was performed using ImageJ as described in Lombardero et al. (2016).

For analyses of S. noctilio body size, egg number and allometric equation development, we incorporated additional data from 1,511 emerging S. noctilio, collected from 53 trees sampled in central NY in 2008; hereafter referred to as the “Central NY data set” (see Zylstra et al. (2010) and Myers et al. (2014) for protocol details). These trees were chemically girdled in 2007 and felled after the oviposition season finished, at which point they were transported to the USDA APHIS PPQ laboratory (North Syracuse, NY) for storage in emergence barrels. Emerging insects were collected and the following measurements were recorded: hind tibial length, prothorax width, mass and egg count. As these trees were chemically girdled rather than naturally attacked, analyses were done separately from 2013–2015 data.

To compare egg sizes between the USA and Spain, we also measured eggs from S. noctilio collected emerging from bolts in Galicia, Spain in 2013–2015, as described in Lombardero et al. (2016). To further compare demographic patterns between the USA and Spain, we reanalyzed data used to build Table 1 in Lombardero et al. (2016), reproducing some parts of that Table and updating it with additional calculations of numbers of eggs laid, based on published estimates (Madden 1974) as described above.

Data analysis was conducted in JMP Pro 13.0 (SAS Institute 2016) and R Version 3.4.3 (R Core Team 2017). All means are reported with standard errors (SE) unless otherwise noted. In cases of unequal variance, we used Welch’s Test to compare means. As we did not sample at all sites in all years, we included a composite variable Site-Year in some analyses.

We used Maximum Likelihood to estimate the number of wasps emerging per tree, number of insects emerging per bolt and the number of nematodes per parasitized S. noctilio adult. For each of these metrics, we evaluated four candidate distributions and compared them via log-likelihood: Poisson, zero-inflated Poisson, negative binomial and zero-inflated negative binomial (R package VGAM; Yee 2015).

We used Chi-square statistics to test for differences in: voltinism between male and female wasps; sex ratio between pine species and across sites; and nematode parasitism among tree species, sites and wasps of different sexes. We used restricted Maximum Likelihood generalized linear mixed models (R package lme4; Bates et al. 2015) to examine relationships between predictors of interest and emergence date, body size and egg length. We calculated P values using the Satterthwaite approximation for degrees of freedom (package lmerTest; Kuznetsova et al. 2017) and selected the best models by comparing AIC values (AICtab in bbmle package; Bolker et al. 2017) unless otherwise noted in results.

We used simple linear regression to test for a relationship between emergence date and body size, analyzing insects emerging from the P. resinosa and P. sylvestris separately. The impacts of sex and tree species on insect body size were examined using 2-way ANOVA with an interaction term. To describe the relationship between body size and S. noctilio egg number, we used the nls function in the R base package to fit a power function to our egg count data from dissected females, then tested for effects of tree species and voltinism on egg number by analyzing the residuals.

Parasitism rates were estimated as the slope of the regression line (with forced intercept = 0) of numbers of emerging parasitoids versus numbers of all emerging insects (siricids plus parasitoids) (Cochran 1977). We then used multiple regression to assess the effect of tree species on parasitism rate.

To examine relationships between bolt traits and measures of insect attack and emergence, we generated a matrix of Pearson Correlation Coefficients using all complete pairwise observations, then visualized them with the R package corrplot (Wei and Simko 2017) using an α = 0.05 significance level. We transformed variables as noted to address departures from normality and heterogeneity of variance. The logit transformations used a standard adjustment (Warton and Hui 2011) of 0.025 (R package CAR; Fox and Weisberg 2011). Patterns in drills per attack were assessed using a generalized linear mixed model with a Poisson link function, that included random effects of tree (nested within site) and bolt (nested within tree). We estimated emergence per attack and emergence per egg as the slope of a regression (with forced intercept = 0) of these pairs of variables (Cochran 1977).

We collected 1007 S. noctilio (253 females and 754 males) over three years from 38 trees harvested in New York, Pennsylvania and Vermont (Table 2). The majority of S. noctilio emergence was concentrated in a small number of trees, with production per tree best described by a zero-inflated negative binomial (ZINB) distribution with Φ = 0.32 (proportion of excess zeroes), µ = 75.39 and overdispersion parameter k = 0.35 (Suppl. material 1: Fig. S3). Most wasps (89.6%) emerged within one year of bolt collection; the rest (10.4%) emerged in year two. Females were more likely than males to emerge in year 2 (21.7% vs. 6.6%; χ² = 40.53, df = 1, p < 0.0001). Woodwasps emerged from 3 July to 2 November (Julian Days 184–306). This timing did not vary across years, but peak emergence date was slightly earlier for males than for females (F_1,839 = 3.74, p = 0.053). Approximately 77% of the variation in emergence dates was among trees within a site.

The sex ratio of emerging S. noctilio was male biased 1:2.98 ♀:♂. Voltinism influenced sex ratios, with insects emerging in one year having a sex ratio of 1:3.5 ♀:♂ and females dominating in year 2, with 1:0.90 ♀:♂. Sex ratio also varied across sites, with the lowest proportion of males in a P. resinosa stand (Blacklick, Pennsylvania) and the highest at a P. sylvestris stand (Delmar, Pennsylvania; 0.32 vs. 0.83; χ² = 59.43, df = 4, p < 0.0001). Although both of these sites are in Pennsylvania, the time since the stands had first become infested appeared to differ: S. noctilio had likely recently arrived to Blacklick, while a population was established in the area around Delmar since at least 2008 (Williams and Hajek 2017).

Sirex noctilio body length ranged from 6–37 mm, with an average length of 15.66 ± 0.16 mm. Females (n = 251, 19.67 ± 0.31 mm) were larger than males (n = 752, 14.32 ± 0.15 mm) (t = 15.40, df = 369.63, p < 0.0001; Fig. 1) and insects emerging in the second year were larger than those emerging in the first (F_1,860.69 = 12.41, p < 0.001). Year of emergence effects were consistent across sexes but more pronounced for males (1^st year: 14.08 ± 0.15 mm vs. 2^nd year: 17.83 ± 0.62 mm; t = 5.92, df = 52.71, p < 0.0001) than for females (1^st year: 19.4 ± 0.37 mm vs. 2^nd year: 20.65 ± 0.56 mm; t = 1.66, df = 104.83, p = 0.03). In the Central NY data set, females ranged from 10–35.5 mm and were larger than males (23.64 ± 0.25 mm vs. 16.95 ± 0.12 mm; F_1,1500 = 239.38, p < 0.0001). Allometric relationships between body size measurements (length, width of head capsule and mass) for insects in the Central NY data set are included in Suppl. material 1 (Suppl. material 1: Fig. S4).

Figure 1. 

Body size variation in Sirex noctilio. Females were larger than males and insects emerging from Pinus resinosa were larger than those emerging from P. sylvestris (Central NY and Northeast data sets plotted separately). Bars show mean ± SE.

Number of eggs per female ranged from 5 to 284. Number of eggs was positively related to body size and well described by a power function where number of eggs = 0.17∙(BodyLength)^2.072 (Fig. 2). Analysis of residuals from the fitted power function showed no effect of tree species or voltinism on number of eggs (F_3,141 = 0.31, p = 0.82). Egg lengths ranged from 0.99–1.51 mm with an average of 1.26 ± 0.01 mm. We found no difference in egg size between Spain and the USA (t = 0.48, df = 21.99, p = 0.64). There was some variation in egg size among wasps, but very little variation among eggs produced by each individual (estimated variation due to random effect in mixed model = 68 and 0%, respectively).

Figure 2. 

Relationship between body size and fecundity. Number of eggs (y) increases with female Sirex noctilio body size at the same rate for the main data set (closed circles) and Central NY data set (open circles). Line is fitted power function y = 0.17x^2.072 for the combined data set.

Dissection of 76 bolts from 24 trees (10 P. radiata and 14 P. sylvestris) yielded 11,253 attacks comprising 16,604 oviposition drill holes. Attack density was higher in P. sylvestris than in P. resinosa (8.83 ± 1.29 vs. 5.28 ± 1.72 attacks/dm²; F_1,51.72 = 4.43, p = 0.04). The number of drills per attack ranged from 1 to 6, distributed as follows: 64% were single drills, 27% doubles, 8% triples and 1% four or more. The number of drills per attack was slightly higher on average for P. resinosa (1.62 ± 0.01) than for P. sylvestris (1.38 ± 0.01) (Fig. 3; n = 11,253, z = -2.21, p = 0.03). The estimated variation in drill count per attack among bolts (nested within tree) and among trees (nested within site), was low (< 10% of the total random variation for each).

Figure 3. 

Number of drills per attack across sampled tree species. Sirex noctilio attacks (oviposition sites) with 1, 2, 3 or 4+ drill holes for Pinus sylvestris and P. resinosa.

Emergence of adult wasps was concentrated in a relatively small number of bolts and was best described by a zero-inflated negative binomial (ZINB) distribution with parameters Φ = 0.24, µ = 20.43 and k = 0.85. For both tree species, emergence was — as expected — positively correlated (r = 0.36 - 0.79) with attacks, attack density, drills, drill density, drills per attack, estimated number of eggs laid and the estimated density of eggs laid, most of which were correlated with each other (Suppl. material 1: Table S1). In P. sylvestris, but not P. resinosa, variables associated with oviposition (estimated eggs laid, egg density), emergence per egg and total emergence were negatively correlated with bolt moisture, bluestain presence and lesion formation in response to attack (Suppl. material 1: Table S1). Lesion formation was positively correlated with bolt moisture level (r = 0.72) in P. sylvestris only. The percentage cross-sectional area of bluestain per bolt was positively correlated with both bolt moisture content and bolt volume (Suppl. material 1: Table S1).

Emergence per attack was higher for P. sylvestris (0.11 ± 0.01) than P. resinosa (0.08 ± 0.01) (Fig. 4); 63.8% of the S. noctilio produced came from the five trees that produced the most insects and these were all P. sylvestris. The proportion of emerging wasps that were male was higher for P. sylvestris (0.80) than for P. resinosa (0.64) (χ² = 28.60, df = 1, p < 0.0001). Insects emerging from P. resinosa were larger than those emerging from P. sylvestris (Fig. 1; Northeast data set: F_1,32.98 = 4.95, p = 0.03, Central NY data set: F_1,1500 = 29.61, p < 0.0001) and the body size difference between males and females was more pronounced in P. resinosa than in P. sylvestris (Northeast only: F_1,974.24 = 8.85, p = 0.003). Variation among trees within species accounted for 32% of the total random variation in adult size.

Figure 4. 

Sirex noctilio emergence success. Emergence per attack for Pinus resinosa (A slope = 0.08 ± 0.01) and P. sylvestris (B slope = 0.11 ± 0.01) and emergence per estimated number of eggs laid for P. resinosa (C slope = 0.16 ± 0.02) and P. sylvestris (D slope = 0.41 ± 0.04). Each point represents one bolt.

A total of 143 native S. nigricornis co-occurred with S. noctilio in our trees, but emergence of this native siricid was concentrated in three trees from one site (Table 2; Suppl. material 1). Hymenopteran parasitoids were abundant: we collected a total of 245 Ibalia leucospoides and 67 rhyssines (27 females and 40 males). Across all sites and years, total hymenopteran parasitism was 23.4 ± 2.0%, with no difference between P. resinosa and P. sylvestris. Ibalia leucospoides parasitized 20 ± 2.0% of S. noctilio larvae, with no difference between tree species. The percentage of S. noctilio parasitized by I. leucospoides increased with total emergence in P. resinosa (F_1,11 = 5.11, p = 0.04), but not in P. sylvestris. Rhyssine parasitism was 3.5 ± 1.0% overall and was higher in P. resinosa (9.22 ± 2.0%) than in P. sylvestris (1.72 ± 0.30%; F_1,34 = 15.57, p < 0.001).

Of the 806 S. noctilio assessed for nematode parasitism, 62% contained nematodes in gonadal tissue, with no difference between sexes (χ² = 0.25, df = 1, p = 0.62). We found no instances of nematodes within S. noctilio eggs. Wasps emerging in the first year were more likely to be parasitized (63.52%) than those emerging in year two (13.64%) (χ² = 22.64, df = 1, p < 0.0001). Among wasps emerging in year one, non-parasitized females were larger than parasitized females (22.03 ± 0.62 mm vs. 17.60 ± 0.49 mm; t = 5.58, df = 158, p < 0.0001) with the same pattern seen in males (t = 4.75, df = 644, p < 0.0001). Females also had smaller eggs in the presence of nematodes (t = 3.17, df = 21.98, p = 0.004). Adult woodwasps that emerged from P. sylvestris were much more likely to be parasitized by D. siricidicola than those that emerged from P. resinosa (81.3% vs. 5.15%; χ² = 363.85, df = 1, p < 0.0001), but species and location were confounded (Table 2), so this finding could simply indicate spatial patchiness in nematode abundance. Estimates of nematode load per parasitized wasp ranged from 11 to 3,000 – the distribution of these values was best described by a negative binomial distribution with parameters µ = 395.2 and k = 1.04. We found many wasps with < 500 nematodes and a few wasps with estimated parasite loads of up to ~ 2,000 or greater.

We constructed simple demographic models of S. noctilio in the northeastern USA, reporting study-wide values in addition to separate values for insects colonizing the two pine species in the study, P. resinosa and P. sylvestris (Table 3). An average-sized female of S. noctilio was predicted to contain 78 eggs (~ 66 eggs when emerging from P. sylvestris or 102 eggs when emerging from P. resinosa), based on body size-fecundity relationships. Nematodes did not enter S. noctilio eggs, so they had no direct impact on egg viability. Estimated emergence per egg was higher for P. sylvestris than for P. resinosa (0.41 ± 0.04 vs. 0.16 ± 0.02; Fig. 4), with an average of 0.28 ± 0.03 overall. From the combination of differences in fecundity and emergence per egg, we estimated 16 and 27 larval progeny per female adult for P. resinosa and P. sylvestris, respectively. A hymenopteran parasitism rate of 23.4%, consistent across tree species, further reduced estimated emergence to 13 and 21 adult progeny per female, of which 36% and 20% were female (Table 3). Thus, the estimated maximum number of females per female per year (λ) was about 4.2 for the overall population or 4.52 and 4.17 for P. resinosa and P. sylvestris, respectively.

Non-native S. noctilio populations in North America showed a higher potential for population growth than native populations in Galicia, Spain (Table 3), with λ > 4 vs. 1.57. This was driven by the absence of sterilization by nematodes in North America and notable differences in fecundity and larval survival inside host trees. Despite this higher potential for population growth, trees with S. noctilio were rare in North America and stems with signs of S. noctilio attack often failed to produce adult progeny. Over 64% of S. noctilio in this study came from 13% of sampled trees, similar to findings from Foelker (2016), where 53.5% of S. noctilio emerged from 16% of sampled trees. Williams and Hajek (2017) and Haavik et al. (2018) also recorded wide variation in emergence and frequent failure of S. noctilio reproduction in attacked trees. Similarly, 32% of trees in this study had no insect emergence, compared to 41% in the native range in Galicia, Spain (Lombardero et al. 2016). Despite increased sampling effort, Lombardero et al. (2016) collected only 313 adults from 134 trees compared to our 1007 from 38 trees.

Our results broaden the evidence of remarkably high body size variation for S. noctilio, consistent with reports of an 8–38 mm range in length from the native range (Lombardero et al. 2016). When compared to other insect species known to have high size variability, female S. noctilio showed roughly twice as much variation (over 4.1× versus 2.1× in the next most variable species; Suppl. material 1: Table S2). This is surprising given the clear link between body size and fecundity (Honěk 1993), although Madden and Coutts (1979) hypothesized that adult size variation is determined by nutrition and the success of the mutualist fungus A. areolatum and other recent studies have supported the hypothesis that highly variable environmental or nutritional factors help determine adult body size (Foelker and Hofstetter 2014; González et al. 2014). This is also consistent with recent findings by Garnas et al. (2020), where a complex interplay of factors related to resource quality influenced woodwasp body size, even within a single tree. Although P. resinosa produced fewer S. noctilio individuals, they were ~ 23% larger than those produced in P. sylvestris, suggesting that P. resinosa is an attractive host, but may have higher host resistance to S. noctilio. When oviposition is successful, P. resinosa provides a suitable and perhaps superior substrate for A. areolatum growth and S. noctilio maturation.

High variation in woodwasp body size could also reflect genetic differences among wasps and complex selective landscapes that favor large insects in some instances, but not in others. Long feeding galleries necessary to produce large adults may be more feasible in large trees than small trees. Large males may have difficulties mating with small females (Caetano and Hajek 2017) and large adults of either sex may be more vulnerable to predators. The large size of female S. noctilio relative to males suggests the effect of natural selection on female size (likely favoring higher fecundity) is considerably stronger than natural or sexual selection on males (Wiklund and Karlsson 1988).

The fecundity of North American S. noctilio is higher than in Spain, but this is not driven by differences in body size. An average-sized female in our sample (19.67 mm long) had ~ 78 eggs. Although Spanish S. noctilio, measured by Lombardero et al. (2016), had a similar range in egg number (0 to 270 eggs per female), 19.7-mm females in Galicia, Spain were estimated to have ~ 56 eggs. Other North American studies also show elevated S. noctilio fecundity compared to native populations (Table 1). On the other hand, fecundity in this study was lower than that reported for Southern Hemisphere populations (e.g., 50–500 in New Zealand; Zondag and Nuttall 1977). The higher fecundity of S. noctilio in the Southern Hemisphere may be attributable to larger average body size, which has been widely reported (e.g., 24–32 mm average body size for females in South Africa; Hurley et al. 2008).

At a population level, observed male-biased sex ratios were similar to those previously reported in North America (see Table 1) and in the native range in Galicia, Spain (Lombardero et al. 2016). In this study, spatial patterns in North American sex ratios were opposite of what would be expected if the male bias were due to low mating success at the range edge since the lowest proportion of males was in the most recently invaded region (Blacklick in central Pennsylvania) where attack densities were low. However, sex ratio has been found to vary widely among sites in other studies (e.g., Haavik et al. 2018) and has the potential to influence local or regional outbreak patterns.

Host choice and oviposition behavior influence the success of S. noctilio in new landscapes. In our study, 64% of total attacks involved only a single drill. Single drills are thought to indicate rejection of the oviposition site (since eggs are rarely placed), perhaps as a consequence of the detection of suboptimal moisture levels and/or oleoresin pressure for fungal establishment, egg survival or larval success (Ryan and Hurley 2012; Hayes et al. 2015). Attacks with two or more drills normally contain eggs, as well as fungi and venom (Madden 1974). In Spain, only 43% of total attacks involved a single drill and, in Argentina, only 39% had one drill (Martinson et al. 2018). Evidently, the US stands in this study included a higher proportion of trees that were judged as low quality by ovipositing females. This may be due, in part, to rejection of trees with higher moisture content (Suppl. material 1: Table S1), variable host species attractiveness and differences in biotic and abiotic drivers of tree stress in different regions (Madden 1988; Haavik et al. 2017; Corley et al. 2018). Any factors that reduce host attractiveness and, thus, oviposition by female woodwasps have the potential to suppress population growth that can lead to outbreaks.

The fate of larvae inside host trees strongly influenced reproductive potential in both US and Spanish S. noctilio populations. This is consistent with past studies that have highlighted the role of host resistance in limiting S. noctilio reproduction and spread (Haavik et al. 2017; Haavik et al. 2018). Mortality at the egg and larval stages consistently reduced numbers of female offspring by over 70% (Table 3, Lombardero et al. 2016) and our finding of much higher reproductive success in P. sylvestris than in P. resinosa further highlights the potential for tree-specific factors to heavily influence S. noctilio population dynamics. However, other demographic factors also matter. For example, the lower survival of larvae feeding in P. resinosa vs. P. sylvestris was almost completely compensated by larger adult size (and, therefore, higher fecundity) and a higher proportion of females (Table 3). In the northeastern United States, small localized outbreaks have occurred, but tended to decline after several years as susceptible host trees were depleted. The potential for woodwasp population growth has probably been highly constrained in the Northeast by the limited availability of suppressed trees within stands of hard pines which are themselves rare in the landscape (Haavik et al. 2016).

Spanish woodwasp populations experience consistent top-down control by nematodes via sterilization of ~ 90% of the eggs in 39% of females (Lombardero et al. 2016). In contrast, sterilization of eggs by parasitic nematodes was absent in our US study populations. Although we found nematodes in ~ 63% of wasps sampled, they were only present in the body cavity, not within the eggs. Previous North American studies have also reported that D. siricidicola fail to sterilize S. noctilio eggs (but see Kroll et al. 2013 and Williams and Hajek 2017 for minor exceptions). Nematode impacts on reproductive potential in the region are, therefore, indirect, but may still be consequential: parasitized females in North America were smaller and less fecund than unparasitized females (17.6 mm and ~ 65 eggs vs. 22.0 mm and ~ 103 eggs). If nematode parasitism reduces adult body size, this corresponds to a 37% drop in the reproductive potential (λ) for parasitized vs. unparasitized females.

Our finding that nematode parasitism dropped to under 14% for larvae that took two years to develop supports the hypothesis that delayed development can help woodwasps evade parasitism (Corley and Bruzzone 2009). The incidence of nematode parasitism was also clumped (better fit by a negative binomial than a Poisson distribution; Vale et al. 2013). Presumably the patchy dispersion of nematodes reflects variation among trees in the introduction of nematodes by ovipositing woodwasps, in the establishment of A. areolatum on which the nematodes feed, and perhaps in the relative timing of wasp and nematode development, which must be synchronized for nematodes to effectively disperse inside emerging woodwasp adults. Even in Southern Hemisphere plantations where D. siricidicola is deployed for biological control, parasitism rates vary widely (Slippers et al. 2012) and annual augmentative releases are regularly practiced and typically necessary for effective control.

Hymenopteran parasitoids exerted a moderate top-down influence on North America S. noctilio, reducing the number of progeny by ~ 20%, which is within the range of reports from other studies in North America (~ 1–50%; Table 1) and the native range. In the most robust study of S. noctilio parasitoid complexes in North America to date, Foelker et al. (2016) found similar overall rates of hymenopteran parasitism (27.6% and 20.9% in 2010 and 2011). We found evidence of density-dependent effects from I. leucospoides at the tree level in P. resinosa. Similarly, Haavik et al. (2016b) found that I. leucospoides presence in Ontario was positively correlated with the density of attacked pines within stands. Ibalia leucospoides uses olfactory cues to locate S. noctilio oviposition sites, based on the presence of volatiles from the woodwasp mutualist A. areolatum (Martínez et al. 2006) and may detect resource-rich patches from some distance (Fischbein et al. 2012). A Type III functional response to S. noctilio presence has also been reported for hymenopteran parasitoids in Argentina, with attack rate increasing as the availability of S. noctilio oviposition sites increases (Fernández-Arhex and Corley 2005). At high S. noctilio densities, this parasitoid can help to suppress outbreaks. However, even if our highest I. leucospoides parasitism rate (33%) was found across all sites, this would only decrease lambda for S. noctilio in the northeastern USA by ~ 17% (Table 3). We suspect that the extreme patchiness of S. noctilio resources in the current North American range, combined with the solitary lifestyle and life history of I. leucospoides (Haavik et al. 2015), limits the potential for population regulation by these parasitoids.

The presence of native pines, native siricids and native siricid parasitoids in North America may confer some biotic resistance to invasion by S. noctilio (Foelker 2016; Nunez-Mir et al. 2017). This differs from the Southern Hemisphere where neither pines nor their associates are native. Despite the similarities to the native range, our demographic analysis suggests higher population growth potential in North America vs. Spain (Table 3). This may result in complex dynamics if the woodwasp reaches native pine forests that are overstocked or experience frequent drought stress, attack by other pests and pathogens or wildfire (Adams et al. 2010; Anderegg et al. 2015). Interactions with other insects will also help shape these dynamics. Our study did not attempt to quantify competition, but we did see reduced S. noctilio emergence in the presence of bluestain fungi associated with bark beetle colonization. This is consistent with other studies where co-occurrence with other subcortical insects has been shown to reduce the reproductive success of S. noctilio, likely mediated by competition among fungal associates (Ryan et al. 2012). These fungal interactions also have the potential to influence woodwasp-nematode interactions and were found to reduce the survival of B. siricidicola and, thus, the effectiveness of biological control efforts, in Australia (Yousuf et al. 2014).

All hard pine species in North America are potentially susceptible to S. noctilio (Dodds and de Groot 2012). Populations in Michigan and southern Ontario colonize P. banksiana and will continue to encounter this host as they spread into new regions. Woodwasps expanding southwards will soon encounter southern pines including P. palustris, P. taeda and P. elliottii and establishment in western North America would permit colonization of lodgepole pine, P. contorta, which has been shown to be highly susceptible to outbreaks in Argentine Patagonia (Lantschner and Corley 2015). More comprehensive knowledge of variation among pine species in susceptibility to S. noctilio would help managers anticipate future impacts in North America.

More broadly, a comprehensive comparison of the population demography of Northern and Southern Hemisphere S. noctilio would help us better elucidate the importance of specific controls on S. noctilio populations and improve understanding of variable tree species susceptibility. Such studies would also help clarify the importance of landscape-level patterns in resource availability in determining woodwasp population growth rates. Current and future population models would also be improved by better understanding and incorporating density-dependent feedbacks in the demographics.