Surveys were conducted on regenerating stands of lodgepole pine to determine the relationship between root deformation and susceptibility to attack by the Warren root collar weevil, Hylobius warreni Wood. The total number of trees attacked by H. warreni did not differ between planted and natural trees. A matched case-control logistic regression suggested that root cross-sectional area was more important in predicting weevil attack for naturally regenerated trees than for planted trees, but weevils were associated with a larger reduction in height-to-diameter ratios for trees with planted root characteristics than for trees with natural root form. Neither the stability of attacked versus unattacked trees differed significantly and there was no significant interaction of weevil attack and tree type, but weevil-killed trees had different root characteristics than alive, attacked trees. Lateral distribution and root cross-sectional area were significant predictors of alive attacked trees versus weevil-killed trees, suggesting that trees with poor lateral spread or poor root cross-sectional area are more likely to die from weevil attack. We conclude that root deformation does not necessarily increase susceptibility to attack but may increase the likelihood of mortality. Thus, measures to facilitate good root form are needed when planting pine in areas with high risk of Warren root collar weevil attack.
The Warren root collar weevil, Hylobius warreni Wood (Coleoptera: Curculionidae), a flightless, long-lived weevil native to the boreal and sub-boreal conifer forests of North America (Cerezke 1994), is widespread throughout the central interior of British Columbia. It has the potential to become an important pest as regeneration of its primary host, lodgepole pine, Pinus contorta Dougl. variety latifolia Engl., increases (Schroff et al. 2006), particularly in the aftermath of a huge outbreak of the mountain pine beetle, Dendroctonus ponderosae Hopkins (Coleoptera: Curculionidae: Scolytinae) (Klingenburg 2008, Kurtz et al. 2008). Lodgepole pine is the dominant conifer in central British Columbia and has been extensively planted over the past decade. Thus, many existing plantations in this area are now at a susceptible age.
Eggs are laid at or on the root collar of major roots, and larvae feed on phloem in the living tree (Cerezke 1994). Development from egg to adult normally takes 2 yr, and extensive girdling can occur on smaller trees. Weevil attacks occur on lodgepole pine of all ages, but are most evident on 5- to 20-yr-old trees because of impacts on growth and mortality. Young trees severely affected by Warren root collar weevil exhibit above-ground growth reduction and decreased root diameter growth (Cerezke 1974), often leading to instability (Cerezke 1994). Significantly, reduced height and diameter increment occurs when girdling by weevil larvae exceeds 60-80% of the circumference of the stem (Cerezke 1994). Despite these growth losses, the insect is not considered a major pest in lodgepole pine plantations, because mortality caused by weevil attack is usually limited (Cerezke 1974).
Even though there is evidence to suggest a possible link between root development problems and attack by insects, there are no quantitative or qualitative data collected from planted trees to directly verify the effect of root development problems on Warren root collar weevil host choice. Hay and Woods (1978) and Graham and Bormann (1966) (cited in Van Eerden 1978) showed that root deformation, as can occur in planted trees (Robert and Lindgren 2006), serves as an impediment to carbohydrate transport toward the tip of the taproot. Root deformation that inhibits translocation of water and nutrients may alter the susceptibility of trees to attack by insects such as the Warren root collar weevil. The research presented here was conducted based on our observations that planted lodgepole pine trees with severe spiral root seemeded more likely to support infestations of weevil larvae. Weevil attack on J-rooted trees also appeared to occur more frequently or to cause more severe damage than on trees with natural root systems. Because it is unlikely that weevil attack can alter root form (Martinsson 1986), our objectives for this study were (1) to assess the effect of root deformation on susceptibility of lodgepole pine to attack by Warren root collar weevil for planted versus naturally regenerated trees; (2) to compare the incidence and damage of weevil on planted and naturally regenerated trees; and (3) to measure the combined impact of weevil and root development problems on tree growth, stability, and mortality of affected trees.
Materials and Methods
Potentially suitable lodgepole pine stands aged 3-10 yr old were identified as described in Robert and Lindgren (2006). Each stand was visited to ensure that both planted and naturally regenerated trees (hereafter referred to as natural or naturals) were present. The presence of mica, vermiculite, or perlite remnants attached to the root system were used as persistent indicators (i.e., these components of the planting mix remain close to the root system even after many years) of planted tree type as these were independent of root morphology.
We collected data from 16 stands (see Robert and Lindgren 2006 for details). The first eight stands were artificially regenerated (hereafter referred to as planted) with some ingress of natural regeneration. Because natural ingress trees were often smaller and younger than the planted trees, a second set of eight stands (fill-planted stands), where the natural trees would be the same size or larger than the planted trees, was identified for study. Table 1 contains a summary of the characteristics for each of the stands surveyed.
Stands 1-8 were primarily planted trees with natural ingress, and stands 9-16 were naturally regenerated sites that were fill planted. The percentage of trees with deformed roots includes only those with moderate or severely deformed roots systems.
Data were collected in systematic plots along transects with randomly chosen points of commencement. A minimum of 4 and a maximum of 20 plots were completed for each stand for a total of 104 plots in the first set of eight stands and 123 in the second set of eight stands.
In each sample plot (3.99-m radius/50-m2 area), trees were randomly excavated until a maximum of the following were found: two trees (one planted and one natural) with current weevil attack and two trees (one planted and one natural) without weevil attack. A total of 562 trees were sampled in this study.
Sample trees were measured for diameter, height, and current and previous year's leader growth. Diameter was measured 30 cm above the soil line to ensure that any swelling present at the base of the tree caused by poor root form did not skew growth rate results. Height-to-diameter ratio was calculated to correct for effects caused by the relatively large differences in tree size over the age classes sampled (see Table 1).
Tree Stability Assessment.
Before excavation, sample trees were tested for stability using a hanging scale measuring up to 25 kg to measure kilograms of force needed to displace the main stem to an angle of 10° from vertical (Burdett 1978). The force was applied at 30 cm from the base of the tree and applied by pulling the tree to the west once only, because after a tree is displaced once, its stability is reduced.
Root Morphology Assessment.
Coarse root structure and root development were assessed through excavation of the woody root system and assessed as in Robert and Lindgren (2006); a brief description of the method follows. Root morphology was described using a number of characteristics including the total number of laterals (variable name: total laterals) and lateral distribution (variable name: sections with laterals). The diameters of laterals >3 mm were summed to give a measure of total root cross-sectional area that is proportional to root size (Lindgren and Örlander 1978). Finally, the root systems were analyzed for the presence of root deformities. Nine categories were assessed on a scale from 0 to 3 (where 0 = no deformity, 1 = low deformity, 2 = moderate deformity, 3 = severe deformity): (1) taproot deformation, (2) J-shaped root formation, (3) spiral root formation, (4) vertically compressed root system, (5) lateral spread (i.e., extent of root clumping to one side or one area), (6) lateral compression (i.e., number of lateral squashed down in the same direction as the taproot), (7) slit plant morphology (the extent to which the lateral root were flattened into a plane), (8) braiding (interwoven root system), and (9) root pairs (two or more pairs of roots growing together). Each root system received a rating for each root characteristic. An overall root form classification (G = good root form or no deformity, L = low deformity, M = moderate deformity, S = severe deformity) was also assigned to each root system (Fig. 1). The overall classification was determined by the root characteristic rating that occurred most often. For example, if a root system had a high number of moderate deformities (2), the overall rating was moderate (M). Only when scores for all characteristics were zero was the root classified as having no deformity.
Root Collar Weevil Damage.
On attacked trees, damage to the root collar was recorded in one of four girdling categories: none (i.e., not attacked by weevil), light (<30% of the root collar circumference girdled), moderate (30-60% girdled), and severe (>60% girdled).
All data analyses were performed on SYSTAT 9, except the mixed model ANOVAs, which were performed on SYSTAT 12 (SYSTAT Software, Richmond CA).
Effect of Root Deformation on Susceptibility to Weevil.
A matched case-control analysis was performed to identify the root parameters that were important in predicting weevil tree-choice. This analysis used matched weevil-attacked and unattacked trees that had been chosen randomly in the same sampling plot. Two separate logistic regression models were created for planted and natural trees to identify the root characteristics that determined weevil tree choice in planted versus natural trees, respectively.
One change had to be made to the data to ensure the absence of multicolinearity, an assumption for logistic regression (Tabachnick and Fidell 2001). All variable correlations were well below 0.7 (as recommended by Tabachnick and Fidell 2001) except for sections with laterals paired with total number of laterals. Sections with laterals was left in the analysis and total number of laterals was removed to avoid violation of this assumption.
Weevil on Planted Versus Natural Trees.
Pearson's χ2 test was used for determining any difference in presence or absence of weevil in planted versus natural trees. The entire data set was pooled for this analysis. The same test was used to uncover any significant differences between the number of planted versus natural trees in each level of weevil girdling damage (low, moderate, and severe girdling damage) and to determine any significant differences between the number of trees with low, moderate, and severe girdling in each root class for planted versus natural trees.
Impact of Weevil and Root Development on Tree Growth Parameters.
Linear mixed model analysis of variance (ANOVA) was used to determine the impact of overall root form and each root characteristic on height-to-diameter ratio of trees. Analysis of covariance (ANCOVA) was used to evaluate any height differences between attacked and unattacked trees for each tree type (planted and natural). Because we were assessing the impact of root form and weevil attack on height, we used diameter as a covariate to account for the size variation among trees. In every ANOVA, stand series was included as a factor in the model.
The data were assessed using Levene's test for homogeneity of variance and a z-statistic for skewness and kurtosis. Where necessary, data were transformed to satisfy the assumptions for ANOVA (Townend 2003). Consequently, height and diameter were transformed as x′ = √(x + 0.5). Two of the deformation categories, vertical compression and root pairs, violated the assumption of equal variances because of the small number of samples showing severe root deformity (rating, 3). To rectify the violation, the rating scale was collapsed into just three ratings (0 = no deformity, 1 = low deformity, 2 = moderate and severe deformity) for these two categories only.
Effect of Root Form on Stability.
Linear mixed model ANOVA was also used to compare stability for attacked versus unattacked trees and planted versus natural tree type. Data were square root-transformed to meet the assumptions of the analysis (Tabachnick and Fidell 2001).
Effect of Root Form on Mortality Caused by Weevils.
A binary direct (where all of the variables are entered into the model simultaneously) logistic regression was used to identify the root characteristics that significantly predicted the differences between live and dead trees with weevil attack. The nine root characteristics (taproot, J-root, spiral root, vertical compression, lateral spread, lateral compression, slit root morphology, braided root, and root pairs) (Robert and Lindgren 2006) in combination with the number of sections with laterals and the average root cross-sectional area were used to predict tree status (live or dead). Total number of laterals was not included in the analysis because the assumption of linearity with the logit was violated and to prevent multicolinearity. Age was included in the analysis to account for the fact that most dead trees were sampled at least a year after their death, so they were consistently younger than the live sampled trees.
Effect of Root Deformation on Susceptibility to Weevil.
The output of the matched case-control logistic regression using all of the individual root characteristics to predict weevil attack (Tables 2 and 3) showed that different models explained some of the varation in weevil attack for both planted (χ2 = 46.640, df = 11, P < 0.001) and natural trees (χ2 = 71.971, df = 11, P < 0.001). Model significance indicated that the root characteristics distinguish between attacked and unattacked trees.
The prediction accuracy and the significant characteristics differed between the planted and the natural models. The planted model (Table 2) suggests that larger values for lateral spread, sections with laterals, and total root cross-sectional are associated with weevil attack. With three significant predictors, the variation in weevil choice explained by the root characteristics model was fairly good (McFadden's ρ2 = 0.323). The prediction accuracy was only moderate, however, at 68.6% correct. None of the odds ratios in the planted model were over two (Table 2).
In contrast to the planted model, only larger root cross-sectional area significantly predicted weevil choice in the natural model (Table 3). Despite this, a much larger portion of the variation in weevil choice was explained in the natural model (McFadden's ρ2 = 0.631). The prediction accuracy (84.2% correct) was also much better than in the planted model. Thus, these results suggest that root characteristics may influence weevil host choice or attack success in planted and natural trees, although the influence of root characteristics may be greater for natural trees.
Weevil on Planted Versus Natural Trees.
The total number of trees attacked by Warren root collar weevil did not differ between planted and natural trees (χ2 = 1.399, df = 1, P = 0.237). The amount of girdling did not vary with root class in planted or natural trees (Fig. 2).
Of the other pests noted on the sample trees, there were more trees in the severe root class with western gall rust (E. harknesii; χ2 = 7.914, df = 2, P = 0.019) and in the moderate root class attacked by pitch nodule moth (Petrova sp.; χ2 = 6.588, df = 2, P = 0.037). There was no significant difference between the number of trees in each root class that were attacked by aphids (Cinara spp.; χ2 = 1.529, df = 12, P = 0.465).
Impact of Weevil and Root Development on Tree Growth Parameters.
Both tree type (F1,333 = 41.508, P < 0.001) and weevil (F1,333 = 45.217, P < 0.001) impacted height-to-diameter ratio for each of the root classes (Fig. 3). Not only was weevil associated with a change in tree shape (height-to-diameter ratio), but weevil attacks tended to occur on the trees with the highest diameter growth rate (F1,316 = 143.329, P < 0.001) and the highest height growth rate (F1,316 = 99.564, P < 0.001; Fig. 4). When diameter was used as a covariate, tree height differed depending on tree type (F1,332 = 6.593, P = 0.011) and the interaction of tree type and weevil (F1,332 = 6.280, P = 0.013; Fig. 5). This interaction was because of a larger difference in height between weevil-attacked planted and natural trees than for unattacked trees.
Effect of Root Form and Weevil on Tree Stability.
There was no significant difference between the stability of natural trees with and without weevils (F2,168 = 1.290, P = 0.258), and there was no significant difference between the stability of planted trees with and without weevils (F2,259 = 0.340, P = 0.561).
Effect of Root Form on Mortality Caused by Weevils.
Direct logistic regression using root characteristics distinguished between weevil-attacked trees that were alive versus those that were dead from weevil (Table 4). This model produced a moderate prediction accuracy of 74.7%. The prediction accuracy was much higher for live trees (83.7%) than for dead trees (43.1%). A moderate value for McFadden's ρ2 value (0.244) indicated that only some of the variation in live and dead trees with weevil was explained by root form.
Three root characteristics reliably distinguished between live and dead trees with weevil were predicted with higher spiral root, higher number of sections with laterals, and higher average cross-sectional area per lateral. Age was also significantly higher for live trees with weevil (Table 4).
Effects of Root Form on Weevil Host Susceptibility.
The shape of the root system may have little direct effect on oviposition choice by adult weevils given that there is no significant difference in the number of attacked trees between planted and natural trees. This is in contrast to Hylobius radicis Buchanan, a similar species in eastern North America, which is more common and severe in plantations and windbreaks than in naturally regenerated stands (Wilson and Millers 1983). Because Warren root collar weevils prefer dominant and co-dominant trees and adult females distribute their eggs at the root collar or in the surrounding soil (Cerezke 1994), below-ground root form may not impact adult host selection.
Even if adult host selection is not directly affected, root form could affect susceptibility of lodgepole pine to weevil in two ways. If root deformation changes above-ground growth parameters, host selection by adults may be indirectly affected. Root form could also alter larval growth or survival. Both the planted and natural model of weevil choice showed highly significant values for total root cross-sectional area. Larvae are often found feeding along the surface of the lateral roots, and a larger root cross-sectional area may therefore mean more resources for the developing larvae (Cerezke 1994). Furthermore, a larger root cross-sectional area could allow the tree increased water and nutrient uptake, which could accelerate growth of the host tree (Hay and Woods 1978). As adults select larger hosts, faster growing trees in a given area would be more susceptible to attack (Cerezke 1994).
Despite the high significance of root cross-sectional area in both models, planted and natural tree models of weevil choice differed in the importance of the root characteristics. None of the other root characteristics affected weevil choice in natural trees, whereas lateral spread and the number of sections with laterals were significant in the planted tree model. Significance of lateral spread and distribution (number of sections with laterals) indicates that even if planted trees have a large surface area, a good lateral spread is also required for a suitable planted host.
Another difference between the planted and natural tree models is the higher amount of variance in weevil choice explained by the root parameters for natural trees. This suggests that there are factors affecting tree host selection or larval survival in planted trees that were not present in the root characteristics model. Chemical cues (such as host volatiles) often influence host selection and oviposition behavior of insects (Bernays and Chapman 1994, Duke and Lindgren 2006, Johnson and Gregory 2006). One possible explanation is that planted trees emit altered chemical cues because of their increased growth rates, different root systems, or genotype. Reduced water uptake can influence the type and amount of chemical cues emitted by a plant (Bernays and Chapman 1994). There is evidence that adults of the Eurasian root weevil, Hylobius abietis L., are able to adjust host selection and feeding in response to host volatiles (Nordlander et al. 1997). The Warren root collar weevil may also be able to adjust host selection in response to changing host volatiles as a result of planted tree root stress.
Effect of Root Form and Weevil on Tree Growth and Growth Rate.
After a tree is attacked, the resulting growth reduction may depend on root characteristics. Trees with good overall root form appear to allocate more energy into height growth and therefore have a larger height-to-diameter ratio (relatively taller and thinner), whereas trees with poorer root form are associated with less height growth and larger diameters (relatively shorter and wider). Weevil-attacked trees have consistently lower height-to-diameter ratios than unattacked trees regardless of root form. It may be that adult weevils oviposit on or near the largest trees and the subsequent larval feeding reduces height growth. The magnitude of this difference, however, is most apparent for trees with good root form. This is consistent with previous research on growth reduction in pines caused by girdling by weevils. Wilson and Millers (1983) showed that the tallest trees showed greater reductions in height than shorter trees in the same stand for infestations of H. radicis.
The height of weevil-attacked planted trees is significantly lower than the height of weevil-attacked natural trees; this may be a result of lateral root characteristics. A higher number and cross-sectional area of lateral roots in the upper part of the soil may allow a young seedling an initial advantage over those trees with the normal taproot and sparser lateral development (Hay and Woods 1978). In addition, an even distribution (i.e., more sections containing laterals) would allow the tree to exploit this upper part of the soil even if translocation is partially cut off by weevil girdling. Therefore, although both planted and natural weevil-attacked trees are associated with reduced growth, planted trees may be at a further disadvantage as a result of root form.
Effect of Root Form and Weevil on Mortality.
Dead trees with weevil attack show different characteristics than trees that are alive with weevil attack; as the number of sections with laterals and root cross-sectional area increases, the probability of predicting a live tree also increases. This means that live trees with weevil attack have better distribution and size of laterals than dead trees. This may be because trees with a good lateral spread and a larger root system would be able to compensate for the loss of water and nutrient uptake from the part of the root system that was cut off because of girdling by weevils. Based on this evidence, it is likely that planted trees have a higher risk of mortality from weevil attack than natural trees. Unfortunately, it was very difficult to reliably distinguish tree type in dead trees, and therefore we were unable to test this prediction. Nevertheless, measures to promote the development of lateral roots in planted lodgepole pine, e.g., root pruning (Khurana 2007) may be advisable in areas at high risk of Warren root collar weevil attack.
Planted tree root form affected the growth and resilience of lodgepole pine trees when they were attacked by weevils. Although the above-ground appearance of planted and natural trees is similar, this study suggests that root form in combination with weevil attack seems to affect growth and possibly tree mortality. Thus, in areas with high risk for Warren root collar weevil damage, measures to improve root form of planted stock may be beneficial.
We thank Canadian Forest Products for financial and logistical support; Mount Begbie Consultants for financial support; H. Giroday, L. Zukewich, and T. Robert for assistance in the field and laboratory; and M. Gillingham and Chris Johns for statistical advice. Three anonymous reviewers provided valuable suggestions on an earlier version of the manuscript. This research was funded by a Discovery Grant from NSERC to B.S.L. and a UNBC entrance scholarship, a NSERC Post Graduate Scholarship, scholarships from the Entomological Society of Canada, and a GREAT Award from Science Council of BC to J.A.R.