Reduction of tree mortality caused by bark beetle attacks is not only important for forestry, but also essential for the preservation of biodiversity and forest carbon sinks in the face of climate change. While bark beetle mass trapping (a “pull” approach) is implemented in practice, few studies exist to estimate its effect. The more complex “push-pull” tactic has, in contrast, been repeatedly tested during the last decade. I analysed published data from 32 experiments in 9 papers published during 2000–2011 on Ips typographus and Dendroctonus ponderosae, to test if there was an overall effect of antiattractant semiochemicals, that is, if treatments reduced the number of attacks on standing trees at the habitat or stand scale. This meta-analysis showed a substantial overall effect size (treatment-control means divided by their SD) of −0.96, with some heterogeneity but little evidence of publication bias. There was no effect of beetle species or publication year. Heterogeneity resulted from different designs and beetle population levels (as year of study). The conventional “% Reduction” measure correlated well with effect size (r2=0.7). Recommendations include more precise reporting of responses (avoiding dichotomous data), more unified experimental designs, and further meta-analyses that include “grey literature” and more beetle species.
1. Introduction
Strategies to reduce tree mortality caused by bark beetle attacks [1–5] are becoming urgent, not only for forest industry, but also for preservation of forest cover and forest carbon sinks in the face of climate change [6–9]. There is a need for quantitative reviews of management alternatives such as mass-trapping and push-pull [10]. While bark beetle mass trapping (a “pull” tactic) is partly implemented in practise, few replicated studies exist to estimate the efficacy of this approach [1–4]. The more complex “push-pull” (as defined by Cook et al. [11]) and “push” (as defined by Gillette and Munson [12]) tactics have, in contrast, been experimentally tested and reported in >10 papers in the last decade, so a review effort on the subject seems to be timely.
I chose the quantitative method of meta-analysis that uses the descriptive data obtained (means, standard deviations, sample sizes), rather than relying on the P values and other analytical statistics which are heavily dependent on sample size when declaring an experiment “significant” or not. Meta-analysis is the method of choice for estimating interventions in complex systems such a clinical medicine, social work, and education [13–16], but also in resolving complex ecological issues [17–20], though less often in applied ecology [10, 21]. In principle, a meta-analysis describes the effect of a treatment among controlled experimental studies by analysing the distribution of effect sizes of the relevant studies [10, 22, 23]. The effect size measure in simple terms is the difference between control and treatment, gauged by the size of their standard deviations, which provides a common scale for the magnitude of effects for the experiments performed [24, 25]. A common “currency” such as the effect size is mandatory for any comparisons between experiments of different designs, sample sizes, and so forth, as measures dependent on sample size like P values cannot be used. Effect sizes can be calculated in several ways; I use the simplest possible often referred to as Cohen’s d [24]:
(1)d=MeanControl-MeanTreatmentSDPooled,
where d above is (1) in [24] and SDPooled is (2) in [24] which takes in account different sample sizes for control and treatment samples.
This meta-analytical review is limited to studies of standing trees at the plot or stand scale, with just two of the recently best studied bark beetle species, Ips typographus and Dendroctonus ponderosae, and published from 2000 and onwards in readily available international journals (thus excluding government reports, trade journals, and other so called “grey literature”). I keep to (1) only newer studies in order to concentrate on the latest development of semiochemicals representing “state of the art” and (2) stand or plot level tests, because they are the only type likely to give information for future practical applications. Tree scale experiments are nowadays essentially a thing of the past and do not provide essential information for forest management. The intended coverage of the literature is not the most comprehensive possible, but is clearly stated and thus open to future challenges.
In this first meta-analysis I hope to address the key question of whether there is any overall effect in reducing or stopping bark beetle attacks on standing trees by antiattractants. Distribution of effect sizes will be scrutinised for evidence of publication bias and heterogeneity. The latter will be studied for meaningful variation among results that may shed light on factors giving low or high effects of semiochemical interventions against beetle attacks.
The intended audience for this paper is not primarily those familiar with the workings of meta-analysis and its application to yet another field, but the colleagues involved in designing and performing field experiments with antiattractants against bark beetle attack on the stand or habitat scale and the forest managers interested in application of such tactics in forestry.
2. Materials and Methods
I have tried to adhere to the standards suggested by PRISMA and Cochrane collaborations [13, 15, 25–28], while acknowledging that these standards are primarily oriented to well-controlled clinical or laboratory studies. The somewhat less rigorous standards of ecological meta-analysis [10, 18, 19] must be used here, in particular to arrive at a sufficient number of experiments to allowing meaningful number for plots and tests for publication bias and sensitivity analysis [29, 30].
Studies were searched using references in published papers, including a review-like paper in the “grey literature” by Gillette and Munson 2009 [12], and “back-tracking” from these by consulting papers citing these first known papers in Google Scholar. This “centrifugal” multidimensional strategy was iterated by expanding to citing papers, and so forth, until no new relevant papers were discovered.
In some cases reports in noninternational sources were difficult to locate and in addition were sometimes partly overlapping with papers later published in international journals. Therefore, I choose to eliminate any studies that were not published in international journals.
A vast majority of recent papers located concerned either the mountain pine beetle of North America, Dendroctonus ponderosae, mostly on ponderosa and lodgepole pine (Pinus ponderosae, P. contorta) or the European spruce engraver, Ips typographus, on Norway spruce (Picea abies). Thus, I decided to concentrate only on these two species, while ignoring the earlier published works on the Southern pine beetle (D. frontalis) and the Western pine beetle (D. frontalis) and a few other spp with one paper each, in order to limit the overall size and heterogeneity of the material.
Numerical data sets were extracted from papers of various designs, by means and SD or frequencies when provided but sometimes from raw or nearly raw data depending on the presentation and were further summarised in MS Excel to avoid pseudoreplication [31–33]. Care was taken to ascertain the true number of replicates but to allow the different data sets to represent cases depending on variation in design, treatment, or time periods (season, year). In bark beetle population dynamics the variation over time and space is overwhelming and I could conclude after reading the experimental papers that multiple experimental outcomes (data points) are in most studies not autocorrelated. I did my best to avoid autocorrelated data, like from time series within a year as in Jakuš et al. [34] by using it as only one data point, or avoided pseudoreplications within experiments by pooling of data [35, 36].
Further analysis and plotting were done by SPSS 19 software package using command syntax (command files available on request) following suggested algorithms [23, 37, 38]. For a meta-analysis, an effect size for each study or experiment must be extracted from the central and secondary moments (means and a measure of variance, respectively) of the distributions of the treatment and control data (details on the data available are given in Table 1). Depending on the reporting and study design, data were often continuous (trees killed or attacked per plot or area, attack density etc.) allowing simple extraction of means and SDs [24, 25], but are sometime dichotomous (data from outcomes that can be divided into two categories), given as total trees killed/attacked versus alive/unattacked per treatment and experiment [39, 40]. In the latter case, a simple transformation of the ln (odds ratio) to Cohen’s d is possible by d=ln(OR)/(π/3) [41] which then allows a comparison of studies with both types of data. This transformation made it possible to use the Cohen’s d also for the dichotomous data.
Input data for meta-analysis with basic effect size and standard error estimates and weights for individual studies.
Expt nr
Author
Year Publ
Year Expt
Location1
Beetle species
Treatment2
Measure
Dicho-tomous
n Ctrl
Effect size (Cohen's d)
Standard error of d
Inverse SE
Z score of d
Standardised residual of d
Weighted d (by Inv. SE)
% Reduction
1
Borden
2006
03
BC
D. ponderosae
Vn 100 u/ha
% mass att.
10
−3.30
0.73
1.38
−2.80
−3.17
−4.54
85%
2
Borden
2006
03
BC
D. ponderosae
Vn + NHV 100 u/ha
% mass att.
10
−2.90
0.68
1.48
−2.30
−2.80
−4.28
83%
3
Jakuš
2011
08
SK
I. typographus
Vn + NHV
Freq tree kill
5
−1.81
0.85
1.17
−0.99
−0.96
−2.13
73%
4
Borden
2006
03
BC
D. ponderosae
NHV 100 u/ha
% mass att.
10
−1.79
0.56
1.79
−0.96
−1.41
−3.20
59%
5
Progar
2005
02
ID
D. ponderosae
Vn
% kill/Suit.
7
−1.68
0.68
1.48
−0.83
−1.01
−2.49
71%
6
Borden
2007
05
BC
D. ponderosae
Vn + tree rem.
Sum trees
Yes
4
−1.59
0.11
9.33
−0.72
−5.50
−14.82
85%
7
Progar
2005
00
ID
D. ponderosae
Vn
% kill/Suit.
7
−1.54
0.66
1.51
−0.66
−0.82
−2.33
83%
8
Jakus
2003
00
PL
I. typographus
Šumava01
Sum trees
Yes
−1.45
0.81
1.23
−0.55
−0.55
−1.78
92%
9
Bentz
2005
01–03
ID, MT
D. ponderosae
Vn “High”
Att. trees/plo
38
−1.35
0.26
3.88
−0.43
−1.36
−5.24
78%
10
Progar
2005
01
ID
D. ponderosae
Vn
% kill/Suit.
7
−1.35
0.64
1.55
−0.42
−0.54
−2.09
68%
11
Gillette
2009
05
CA
D. ponderosae
Vn arial appl.
Att. trees/ha
5
−1.27
0.76
1.27
−0.33
−0.34
−1.61
69%
12
Jakus
2003
01
CZ
I. typographus
Šumava03
Sum trees
Yes
−1.16
0.28
3.63
−0.19
−0.58
−4.20
84%
13
Jakus
2003
00
CZ
I. typographus
Šumava02
Sum trees
Yes
−1.02
0.60
1.66
−0.02
−0.03
−1.69
83%
14
Borden
2003
01
BC
D. ponderosae
High Vn
Att. trees/plot
10
−0.94
0.50
2.01
0.07
0.12
−1.89
57%
15
Gillette
2009
05
ID
D. ponderosae
Vn arial appl.
Att. trees/ha
5
−0.94
0.76
1.32
0.07
0.08
−1.24
65%
16
Borden
2006
03
BC
D. ponderosae
Vn + NHV 400 u/ha
Freq tree kill
10
−0.93
0.50
2.01
0.09
0.15
−1.86
59%
17
Schiebe
2011
06
SK
I. typographus
Vn + NHV × 4
Prop tree kill
9
−0.88
0.67
1.50
0.14
0.18
−1.32
56%
18
Borden
2006
03
BC
D. ponderosae
Vn + NHV 100 u/ha
Freq tree kill
10
−0.87
0.49
2.02
0.15
0.25
−1.77
51%
19
Borden
2006
03
BC
D. ponderosae
Vn + NHV 178 u/ha
Freq tree kill
10
−0.85
0.49
2.03
0.18
0.30
−1.73
55%
20
Schiebe
2011
07
SK
I. typographus
Vn + NHV × 2
Prop tree kill
9
−0.82
0.66
1.51
0.22
0.28
−1.23
49%
21
Schiebe
2011
07
SK
I. typographus
Vn + NHV × 4
Prop tree kill
9
−0.76
0.66
1.51
0.29
0.36
−1.15
63%
22
Borden
2007
05
BC
D. ponderosae
Vn
Sum trees
Yes
3
−0.66
0.13
7.52
0.41
2.54
−4.96
55%
23
Jakuš
2003
00
CZ
I. typographus
Šumava01
Sum trees
Yes
−0.65
0.17
5.94
0.42
2.07
−3.86
52%
24
Schiebe
2011
06
SK
I. typographus
Vn + NHV × 1
Prop tree kill
9
−0.63
0.65
1.53
0.45
0.56
−0.97
34%
25
Borden
2003
01
BC
D. ponderosae
High Vn + NHV
Att. trees/plot
10
−0.44
0.48
2.09
0.68
1.16
−0.92
23%
26
Borden
2003
01
BC
D. ponderosae
Low Vn
Att. trees/plot
10
−0.40
0.48
2.09
0.72
1.25
−0.84
23%
27
Borden
2006
03
BC
D. ponderosae
Vn + NHV 114 u/ha
Freq mass att.
10
−0.39
0.48
2.10
0.74
1.27
−0.82
38%
28
Schiebe
2011
06
SK
I. typographus
Vn + NHV × 2
Prop tree kill
9
−0.32
0.64
1.56
0.82
1.05
−0.50
40%
29
Schiebe
2011
07
SK
I. typographus
Vn + NHV × 1
Prop tree kill
9
−0.27
0.64
1.56
0.88
1.13
−0.43
47%
30
Borden
2003
01
BC
D. ponderosae
Low Vn + NHV
Att. trees/plot
10
−0.27
0.47
2.11
0.89
1.53
−0.57
18%
31
Progar
2005
04
ID
D. ponderosae
Vn
% kill/Suit.
7
0.02
0.58
1.72
1.23
1.74
0.03
−2%
32
Progar
2005
03
ID
D. ponderosae
Vn
% kill/Suit.
7
1.26
0.64
1.57
2.74
3.55
1.97
−396%
1BC: British Columbia, Canada, SK: Slovak Republic, ID: Idaho, USA, MT: Montana, USA, CA: California, USA, CZ: Czech Republic.
2Vn: verbenone, NHV: nonhost volatiles (combination of green leaf volatiles and bark compounds), Šumava01–Šumava03: dispensers with Vn + NHV.
A random effect model was used, as there was no reason to assume that all studies were functionally identical, as in a fixed model where only a single effect plus error is estimated [38]. Instead the random model includes also variation due to heterogeneity, like different species and phases of the population dynamics were likely to influence results [37, 38] which lowers the magnitude of the estimated overall effect size (θ) and widens its associated confidence interval. Bootstrap re-sampling was not used.
Heterogeneity in recorded effect size distributions may stem from several sources, including true biological variation as well as variation in study designs. Overall for the meta-analysis I followed the results and recommendations of Bax et al. [29] for use of more readily understandable plots for the applied ecologist in the assessment of reporting bias and heterogeneity, rather than the more arcane statistics of QT, I2, τ2, and so forth, [13, 22, 30, 42, 43]. Sensitivity plots and subgroup analysis as well as attempts of metaregression, based on data of different quality, origin, and treatment levels, respectively, were included to explore heterogeneity and deviations from normality [13].
An important consideration for any review is the “file drawer problem” where a publication bias usually exists against the publication of negative outcomes with low effects. A notable exception is the combination of failures and successes by Jakuš et al. [40]. The subject of this paper, in particular may suffer from this problem, as applied experiments with unclear or negative results may often either not be submitted for publication (stays in the file drawer) or get hidden in the “grey literature” of various internal reports, trade journals, and so forth. Fortunately, there exists graphical tools such as “funnel plots” to aid in detection of such publication biases [29].
3. Results3.1. Effect Sizes
The analysis included 32 data points from 9 papers [35, 36, 39, 40, 44–48], where several papers reported experiments from different years or seasons and/or different semiochemical blends [36, 40, 47, 48] which were considered independent studies (Table 1).Overall, the effect sizes fell below zero, meaning that there was a lower density or number of attacks on trees in the treated plots, with only one exception (Figure 1). The effect size confidence intervals for individual studies, however, included zero for all but 12 cases (Figure 1). The overall estimate of effect was θ=-0.96±0.25 (the midpoint and ± the 95% CI of grand mean effect size based on a random model, SE{θ}=0.126). There was a narrow confidence interval for the grand mean compared to those of the individual studies (Figure 1, lowest point, “diamond”). The effect sizes followed in general a smooth progression when ordered after size from low to high, while one study stood out with a reversed (d>0) effect and two with large values (|d|>2) (Figure 1).
Ordered plot of individual study effect sizes and their ±95% confidence intervals; known as a “forest plot”. Vertical lines: — estimated overall effect size (θ), - - - effect size = 0 or no effect. The “diamond” on last row shows the midpoint and width of the 95% CI of grand mean effect size by its location and width, based on a random model estimate [37, 38]. Numbers on the dependent axis correspond to the “Expt nr” in Table 1.
3.2. Sensitivity Analysis3.2.1. Publication Bias
This kind of field research, with highly variable forest conditions and bark beetle populations, and many factors outside the control of authors, might result in “failed studies” (negative or unclear results) that tend to remain unreported. An example to the contrary is the study of Jakuš et al. [40], reporting a string of “failures” before successes. Unfortunately, several of these “failures” could not be included here because they incorporated felled trees or logs, or did not provide sufficient details for estimation of d values. Still, the “funnel plot” (Figure 2) is reasonably symmetrical. If a strong bias existed against publication of unsuccessful studies, this plot would give very few points to the right of the mean d estimate of ≈−1 and no points at or above d=0. Interestingly, “good” studies with high n values or high inverse SEd values do not have a clear bias to the left of d≈-1, but have a rather symmetrical distribution (Figure 2).
Funnel plot for dissemination or publication bias detection. — estimated overall mean effect size. Points distributed symmetrically around the mean effect size indicate little evidence of dissemination bias [29].
There was no effect whatsoever of publication year on effect size (R2=0.002).
3.2.2. Heterogeneity
The plots in Figure 3 indicate an overall moderate heterogeneity [29]. In Figure 3(a), the box plot of d values weighted by their inverse SE, there is one extreme value (from [39]) and one outlier (from [47]). The median d (middle line of box) falls lower than the mean d. The histogram of standardized residuals, Figure 3(b), shows no clear deviations from the normal distribution overlay (indicating a normal sample distribution) or any clumping of values indicating possible subpopulations (indicating little heterogeneity in the sample). Similarly, the normal quantile plots indicate a normal distribution (points close to a straight line) and little heterogeneity (no clustering of points).
Plots to check normality and heterogeneity of effect size in the whole data set (n=32) [29]. (a) Weighted box plot (data points used weighted by division of SE, vertical line is median, box enclose 50% of observations, whiskers 2 SD or 95% of observations, “o” are outliers not enclosed by whiskers, and “*” extremes), where asymmetry indicates heterogeneity and deviations from normality. Numbers near points correspond to the “Expt nr” in Table 1. (b) Histogram of residuals (data-grand mean)/SD allowing check of normality. (c) Normal Quantile Plot of Z-scores (normalised data by subtraction of mean and division by SD), allowing check of normality and of outliers contributing to heterogeneity.
Similarly, there was no variation between the effect sizes based on studies providing continuous data (n=26) or dichotomous data (n=6) in effect of antiattractants (continuous data d, mean ± SE [median] = −0.98 ± 0.18 [−0.88], dichotomous ditto = −1.09 ± 0.16 [−1.09]; F1,31=0.08, P≫10%).
3.3. Subgroup Analysis
There was little variation between the two species in overall effect of antiattractants ( Ips mean ± SE [median] = −0.89 ± 0.14 (−0.82), Dendroctonus ditto = −1.06 ± 0.21 [−0.94]; F1,31=0.3, P≫10%).
From a practical point it would be of considerable interest to know if verbenone alone is as efficient as a blend of verbenone combined with sometimes more expensive nonhost volatiles (NHV); however, the studies are not balanced in this respect. All Ips studies included an NHV blend plus verbenone, as the combination is known to be clearly needed [5, 40], but for Dendroctonus only one study included an NHV blend [36].
3.4. Meta-Regression3.4.1. Year of Study
There was a relatively strong effect of year since the start of study (ANOVA, full corrected model: F6,9=11.0, P=3.8%, η2=0.96) which most likely corresponds to an effect of beetle population variation between years (factor year-of-experiment: F4,9=16.0, P=2.3%, η2=0.96). However, there was also some variation due to factor species (F1,9=5.4, P=10.2%, η2=0.64) but there was no interaction due to year-of- experiment × species (F1,9=0.1, P≫10%, η2=0.24).
3.4.2. Comparison to the Conventional “% Reduction” Measure
The conventional measure of inhibitory effect, “% Reduction”:(2)% Reduction=Attacks(kills)incontrolgroup-Attacks(kills)intreatmentgroupAttacks(kills)incontrolgroup⋅100
is used in papers on reduction of trap catches, attack densities, or tree kills and is common in applied entomology. However, the stated “% reduction” probably has a wide SE (as it includes the subtraction of two variables both estimated with uncertainty, and then divided by one of them), but I have seen no attempts to quantify this uncertainty. Thus, there is a strong incentive to get a relation to the effect size and its well-defined uncertainty. “% Reduction” was regressed upon effect size and corresponded, as expected, overall rather well (R2=0.68) with effect size (Figure 4(a)), but only when omitting the three points that were >2 SDs away from the mean (the three deviates from line in Figure 3(c)). There was little variation between the slopes or the strength of correlation (R2) based on studies providing continuous data or dichotomous data (Figure 4(a)). The Ips data points, which had a smaller spread along the independent axis (x-axis), had a weaker correspondence (R2=0.58) than those for Dendroctonus (R2=0.79) (Figure 4(b)).
Correspondence between effect size and a standard measure of effect magnitude in applied entomology “% Reduction” ([attacked in control − attacked in treatment/attacked in control] * 100) for (a) different data types and (b) the two different species (Ips/Dendroctonus).
4. Discussion
In applied ecological applications such as forest entomology, critical reviews in the form of meta-analysis are not wide-spread [10] but see [7, 20, 21]. Sometimes titles of reviews may imply similarity in scope but, due to criteria set for included studies, have little overlap, as in [42], which covers semiochemical manipulation of pest insects but not insect pest management per se. Interestingly, the paper by Szendrei and Rodriguez-Saona [42] does not show a single experimental paper in common with the present review.
In principle, a meta-analysis describes the effect of a treatment among controlled experimental studies by analysing the distribution of effect sizes of the relevant studies [10, 22, 23]. By using results of many studies, it may allow detection of effects not easy to confirm by single studies [49].
The literature was screened by a “centrifugal” multidimensional strategy using citations to some core papers and then iterated by expanding to citing papers, and so forth, until no new papers turned up. This manual approach, based on my knowledge of the field, and the omission of a traditional one-dimensional database search step was justified by the finding that a single data base search may locate only 1/3 of relevant studies [50]. Back-tracking might have created an “inbreeding effect” of focus on papers by a possible citing bias, but I found no clear such indication and would leave this problem for a future discussion based on more studies.
Heterogeneity in effect size may come from several sources, including both true biological variation such as beetle population size and tree vigour in the field, as well as variation in study design and treatment levels. Some papers present details of “different” experiments that are more of replications of the same design rather than different experiments, while others join or would allow joining of data of similar experiments conducted in different years under different conditions [47, 48]. A clear example is that of Progar [47] who clearly showed a reasonable variation in effect over the years of an outbreak, where in the beginning treated plots were protected from attacks by an escalating beetle population, but later became the victims of being the only remnants of mature trees left in the landscape, suitable to be attacked by the beetles at the peak of the outbreak [51].
An important consideration for any metastudy is the “file drawer problem” where a publication bias usually exists against the publication of negative outcomes with low effects. A notable exception is Jakuš et al. [40]. For the current analysis in particular, this aspect is an important consideration as the “grey literature” is not covered, which may cover some less successful experiments. The analysis of effect size distribution in this meta-analysis did, however, not detect any clear pattern of publication or dissemination bias.
The development of complete anti-attractant (verbenone) technology for the Southern Pine Beetle Dendroctonus frontalis was successfully finished in 1990s. The evaluation of treatments was based on growth rates of infestations. Clarke et al. [52] has shown that verbenone-only tactic completely suppressed 69% of infestation rate and verbenone-plus-felling tactic suppressed 86%, corresponding to an effect size of |d|>1 and >1.5, respectively, based on the regression in this paper (Figure 4). If one may rely on the regression of % Reduction on effect size in this paper and on these higher numbers for D. frontalis as representative, it is likely that the development of “push and pull” technology for I. typographus and D. ponderosae has some promise for future development. There is in my view a need for more complex anti-attractant mixture for both I. typographus and D. ponderosae (push) and for combinations with pheromone traps (pull) [34, 36, 40, 53–55]. It could be interesting to note that an effect size grand mean estimate in this paper θ≈-1 corresponds to a reduction of attacks of about 60%, a value found already in 2003 by Jakuš et al. [40] on data from 2000 and 2001 for Ips.
Patterns and mechanisms of the response by the tree to beetle attack are well researched in Dendroctonus [51, 56] and information is now becoming available for Ips [57, 58]. Contrarily, beetle response to signals from tree or beetle is in recent years better covered in Ips, especially at the antennal and single sensillum level of peripheral detection [59–62].
A deeper understanding of the olfactory interface with the environment for D. ponderosae will hopefully emerge from on-going study of the recently published transcriptomes [63, 64] and genome (Keeling et al., unpublished) and the transcriptome of antenna compared for the two species in progress (Andersson et al., unpublished). The present stage of analysis indicates a number of closely related pairs of antennal olfactory receptor genes (OR) that are not found in the other genomic coleopteran Tribolium [60]. Such understanding will in the future help in the study and manipulation of beetle olfaction and resulting behaviours.
This meta-analysis sets the basis that will further the development of antiattractants by quantitatively establishing a substantial overall effect of such interventions in the two species recently most studied. Such confirmed efficacy of antiattractants against attacks by these two “aggressive” bark beetles might encourage better designed and larger-scale applied studies on these and other economically important conifer bark beetle species in the future. The semiochemical-based “push-pull” tactic has played and will continue to play an important role in the pest management practices in agricultural, forestry, medical and urban settings [11]. Future meta-analysis and experimental studies on the subject would include more new study data points (more species) and “grey literature”, but also better reporting (clear provision of means and SDs [65], avoiding dichotomous data) of experimental responses and more of novel designs. Would it, for instance, be possible to use lower densities of dispensers, but with higher release rates as indicated by the distance effects noted in one study [48]?
When anti-attractant technology and treatment designs once comes of age, we will still have to keep in mind that we are not “protecting trees” in reality, but rather helping them to protect themselves. If we design forest landscapes for biodiversity and semiochemical diversity [11, 20, 66], we will in the long run help trees even better.
Acknowledgments
For a constructive critique on language and statistics, respectively, the author thank two anonymous referees. For critical comments on earlier drafts he thank C. Schiebe, Q.-H. Zhang (Sterling International, Inc., Spokane, USA), and R. Jakuš (Inst. Forest Ecology, SAV, Zvolen, Slovak Republic). For analysis, support was provided by Department of Plant Protection Biology, SLU and by Linnaeus-program “Insect Chemical Ecology, Ethology, and Evolution” (ICE3).
DimitriL.GebauerU.LösekrugR.VaupelO.Influence of mass trapping on the population dynamic and damage-effect of bark beetles19921141–5103109FaccoliM.StergulcF.Damage reduction and performance of mass trapping devices for forest protection against the spruce bark beetle, Ips typographus (Coleoptera Curculionidae Scolytinae)20086530919GrégoireJ.-C.EvansH. F.LieutierF.DayK. R.BattistiA.GrégoireJ.-C.EvansH. F.Damage and control of Bawbilt organisms—an overview2004Amsterdam, The NetherlandsSpringer1937WermelingerB.Ecology and management of the spruce bark beetle Ips typographus—a review of recent research20042021–367822-s2.0-634429460510.1016/j.foreco.2004.07.018ZhangQ. H.SchlyterF.Olfactory recognition and behavioural avoidance of angiosperm nonhost volatiles by conifer-inhabiting bark beetles2004611192-s2.0-124232253110.1111/j.1461-9555.2004.00202.xBentzB. J.RgnireJ.FettigC. J.HansenE. M.HayesJ. L.HickeJ. A.KelseyR. G.NegrnJ. F.SeyboldS. J.Climate change and bark beetles of the western United States and Canada: direct and indirect effects20106086026132-s2.0-7795641505810.1525/bio.2010.60.8.6JactelH.PetitJ.Desprez-LoustauM.-L.DelzonS.PiouD.BattistiA.KorichevaJ.Drought effects on damage by forest insects and pathogens: a meta-analysis20121812672762-s2.0-8405518482110.1111/j.1365-2486.2011.02512.xSchlyterP.StjernquistI.BärringL.JönssonA. M.NilssonC.Assessment of the impacts of climate change and weather extremes on boreal forests in northern Europe, focusing on Norway spruce200631175842-s2.0-33746410948SchlyterF.WitzellJ.Forests: carbon storages in peril20093136137StewartG.Meta-analysis in applied ecology20106178812-s2.0-7744909816410.1098/rsbl.2009.0546CookS. M.KhanZ. R.PickettJ. A.The use of push-pull strategies in integrated pest management2007523754002-s2.0-3384638827710.1146/annurev.ento.52.110405.091407GilletteN. E.MunsonA. S.HayesJ. L.LundquistJ. E.Semiochemical sabotage: behavioral chemicals for protection of western conifers from bark beetles2009Portland, Ore, USAUnited States Department of Agriculture, Forest Service, Pacific Northwest Research Station85127BownM. J.SuttonA. J.Quality control in systematic reviews and meta-analyses20104056696772-s2.0-7864945396510.1016/j.ejvs.2010.07.011GlassG. V.McGawB.SmithM. L.198156Beverly Hills, Calif, USASageSchünemannH. J.OxmanA. D.VistG. E.Interpreting results and drawing conclusions2008Chichester, UKJohn Wiley and Sons359387SuttonA. J.HigginsJ. P. T.Recent developments in meta-analysis20082756256502-s2.0-3954909310710.1002/sim.2934Chaplin-KramerR.O'RourkeM. E.BlitzerE. J.KremenC.A meta-analysis of crop pest and natural enemy response to landscape complexity20111499229322-s2.0-7996117385210.1111/j.1461-0248.2011.01642.xGurevitchJ.HedgesL. V.Statistical issues in ecological meta-analyses1999804114211492-s2.0-0033040602LajeunesseM. J.Achieving synthesis with meta-analysis by combining and comparing all available studies2010919256125642-s2.0-7795659433210.1890/09-1530.1JactelH.BrockerhoffE. G.Tree diversity reduces herbivory by forest insects20071098358482-s2.0-3454741444210.1111/j.1461-0248.2007.01073.xShrivastavaG.RogersM.WszelakiA.PantheeD. R.ChenF.Plant volatiles-based insect pest management in organic farming20102921231332-s2.0-7795089171510.1080/07352681003617483PigottT. D.2012New York, NY, USASpringerBorensteinM.HedgesL. V.HigginsJ. P. T.RothsteinH. R.2009Chichester, UKJohn Wiley and SonsNakagawaS.CuthillI. C.Effect size, confidence interval and statistical significance: a practical guide for biologists20078245916052-s2.0-3454790007910.1111/j.1469-185X.2007.00027.xDeeksJ. J.HigginsJ.AltmanD. G.Analysing data and undertaking meta-analyses2008Chichester, UKJohn Wiley and Sons243296ThorlundK.WalterS. D.JohnstonB. C.FurukawaT. A.GuyattG. H.Pooling health-related quality of life outcomes in meta-analysis—a tutorial and review of methods for enhancing interpretability201123188203LiberatiA.AltmanD. G.TetzlaffJ.MulrowC.GøtzscheP. C.IoannidisJ. P. A.ClarkeM.DevereauxP. J.KleijnenJ.MoherD.The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration20096210e1e342-s2.0-76549089252MoherD.TetzlaffJ.LiberatiA.AltmanD. G.The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration2009672-s2.0-6804913786910.1371/journal.pmed.1000100e1000100BaxL.IkedaN.FukuiN.YajuY.TsurutaH.MoonsK. G. M.More than numbers: the power of graphs in meta-analysis200916922492552-s2.0-5844909228410.1093/aje/kwn340HufW.KalcherK.PailG.FriedrichM. E.FilzmoserP.KasperS.Meta-analysis: fact or fiction? How to interpret meta-analyses20111231882002-s2.0-7995444629110.3109/15622975.2010.551544RyanM. J.Replication in field biology: the case of the frog-eating bat20113346060122912302-s2.0-8275517412110.1126/science.1214532HurlbertS. H.The ancient black art and transdisciplinary extent of pseudoreplication200912344344432-s2.0-7294911296110.1037/a0016221HurlbertS. H.Pseudoreplication and the design of ecological field experiments1984542187211JakušR.BlaženecM.VojtěchO.Use of anti-attractants in specific conditions of protected areas201138146512-s2.0-83355169290BentzB. J.KegleyS.GibsonK.ThierR.A test of high-dose verbenone for stand-level protection of lodgepole and whitebark pine from mountain pine beetle (Coleoptera: Curculionidae: Scolytinae) attacks2005985161416212-s2.0-27544475472BordenJ. H.BirminghamA. L.BurleighJ. S.Evaluation of the push-pull tactic against the mountain pine beetle using verbenone and non-host volatiles in combination with pheromone-baited trees20068245795902-s2.0-33746898223DeeksJ. J.HigginsJ. P. T.(2010, 2012-01-15). Statistical algorithms in Review Manager 5, http://ims.cochrane.org/revman/documentation/Statistical-methods-in-RevMan-5.pdfBorensteinM.HedgesL. V.HigginsJ.RothsteinH. R.A basic introduction to fixed-effect and random-effects models for meta-analysis20101297111BordenJ. H.SparrowG. R.GervanN. L.Operational success of verbenone against the mountain pine beetle in a rural community20073353183242-s2.0-34648833602JakušR.SchlyterF.ZhangQ. H.BlaženecM.VaverčákR.GrodzkiW.BrutovskýD.LajzováE.TurčániM.BengtssonM.BlumZ.GregoiréJ. C.Overview of development of an anti-attractant based technology for spruce protection against Ips typographus: from past failures to future success200376489992-s2.0-004324084610.1046/j.1439-0280.2003.03020.xChinnS.A simple method for converting an odds ratio to effect size for use in meta-analysis20001922312731312-s2.0-0034736417SzendreiZ.Rodriguez-SaonaC.A meta-analysis of insect pest behavioral manipulation with plant volatiles201013432012102-s2.0-7764909298510.1111/j.1570-7458.2009.00954.xRückerG.SchwarzerG.CarpenterJ. R.SchumacherM.Undue reliance on I2 in assessing heterogeneity may mislead200881, article 792-s2.0-6164909867410.1186/1471-2288-8-79BordenJ. H.ChongL. J.EarleT. J.HuberD. P. W.Protection of lodgepole pine from attack by the mountain pine beetle, Dendroctonus ponderosae (Coleoptera: Scolytidae) using high doses of verbenone in combination with nonhost bark volatiles20037936856912-s2.0-0141523021GilletteN. E.ErbilginN.WebsterJ. N.PedersonL.MoriS. R.SteinJ. D.OwenD. R.BischelK. M.WoodD. L.Aerially applied verbenone-releasing laminated flakes protect Pinus contorta stands from attack by Dendroctonus ponderosae in California and Idaho20092575140514122-s2.0-5934910619510.1016/j.foreco.2008.12.017JakusR.ZajíčkovaL.CudlínP.BlaženecM.TurčaniM.JežíkM.LieutierF.SchlyterF.Landscape-scale Ips typographus attack dynamics: from monitoring plots to GIS-based disturbance models201142562612-s2.0-8375517894110.3832/ifor0589-004ProgarR. A.Five-year operational trial of verbenone to deter mountain pine beetle (Dendroctonus ponderosae; Coleoptera: Scolytidae) attack of lodgepole pine (Pinus contorta)2005346140214072-s2.0-29544439507SchiebeC.BlaženecM.JakušR.UneliusC. R.SchlyterF.Semiochemical diversity diverts bark beetle attacks from Norway spruce edges2011135107157862-s2.0-79953258706MarvierM.Using meta-analysis to inform risk assessment and risk management20116supplement 1s113s1182-s2.0-7995587078710.1007/s00003-011-0675-6McManusR. J.WilsonS.DelaneyB. C.FitzmauriceD. A.HydeC. J.TobiasR. S.JowettS.HobbsF. D. R.Review of the usefulness of contacting other experts when conducting a literature search for systematic reviews19983177172156215632-s2.0-0032488342BooneC. K.AukemaB. H.BohlmannJ.CarrollA. L.RaffaK. F.Efficacy of tree defense physiology varies with bark beetle population density: a basis for positive feedback in eruptive species2011416117411882-s2.0-7995829515910.1139/x11-041ClarkeS. R.SalomS. M.BillingsR. F.BerisfordC. W.UptonW. W.McClellanQ. C.DaluskyM. J.A scentsible approach to controlling southern pine beetles. Two new tactics using verbenone199997726312-s2.0-0032882674AnderssonM. N.BinyameenM.SchlyterF.Attraction modulated by spacing of pheromone components and anti-attractants in a bark beetle and a moth: a smaller scale of action in the moth2011378899911FettigC. J.McKelveyS. R.DabneyC. P.BorysR. R.HuberD. P. W.Response of Dendroctonus brevicomis to different release rates of nonhost angiosperm volatiles and verbenone in trapping and tree protection studies200913321431542-s2.0-6044911521310.1111/j.1439-0418.2008.01317.xHuberD. P. W.BordenJ. H.Protection of lodgepole pines from mass attack by mountain pine beetle, Dendroctonus ponderosae, with nonhost angiosperm volatiles and verbenone20019921311412-s2.0-003503668510.1023/A:1018979916210RaffaK. F.BerrymanA.Physiological differences between lodgepole pines resistant and susceptible to the mountain pine beetle and associated microorganisms198211486492SchiebeC.HammerbacherA.BirgerssonG.Inducibility of chemical defences in Norway spruce bark is correlated with unsuccessful mass attacks by the spruce bark beetle2012170118319810.1007/s00442-012-2298-8ZhaoT.KrokeneP.HuJ.ChristiansenE.BjörklundN.LångströmB.SolheimH.Borg-KarlsonA.-K.Induced terpene accumulation in Norway spruce inhibits bark beetle colonization in a dose-dependent manner2011610e266492-s2.0-8005478420910.1371/journal.pone.0026649AnderssonM. N.LarssonM. C.BlaženecM.JakušR.ZhangQ. H.SchlyterF.Peripheral modulation of pheromone response by inhibitory host compound in a beetle201021319333233392-s2.0-7795684393310.1242/jeb.044396AnderssonM. N.BengtssonJ. M.Grosse-WildeE.Olfactory receptors in Ips typographus: transcriptome from antenna analysed and preliminary compared to Dendroctonus ponderosae and Tribolium castaneum (Coleoptera: Curculionidae & Tenebrionidae)Proceedings of the Genetics of Bark Beetles and Associated Microorganisms2011Sopron, HungaryAnderssonM. N.LarssonM. C.SchlyterF.Specificity and redundancy in the olfactory system of the bark beetle Ips typographus: single-cell responses to ecologically relevant odors20095565565672-s2.0-6734914708710.1016/j.jinsphys.2009.01.018AnderssonM. N.Mechanisms of odor coding in coniferous bark beetles: from neuron to behavior and application201220121414957210.1155/2012/149572AwT.SchlauchK.KeelingC. I.YoungS.BearfieldJ. C.BlomquistG. J.TittigerC.Functional genomics of mountain pine beetle (Dendroctonus ponderosae) midguts and fat bodies2010111, article e2152-s2.0-7795038208510.1186/1471-2164-11-215KeelingC. I.HendersonH.LiM.YuenM.ClarkE. L.FraserJ. D.HuberD. P. W.LiaoN. Y.Roderick DockingT.BirolI.ChanS. K.TaylorG. A.PalmquistD.JonesS. J. M.BohlmannJ.Transcriptome and full-length cDNA resources for the mountain pine beetle, Dendroctonus ponderosae Hopkins, a major insect pest of pine forests201242852553610.1016/j.ibmb.2012.03.010GripenbergS.MayhewP. J.ParnellM.RoslinT.A meta-analysis of preference-performance relationships in phytophagous insects20101333833932-s2.0-7794989283910.1111/j.1461-0248.2009.01433.xZhangQ. H.SchlyterF.Redundancy, synergism, and active inhibitory range of non-host volatiles in reducing pheromone attraction in European spruce bark beetle Ips typographus200310122993102-s2.0-003804763710.1034/j.1600-0706.2003.111595.x