Interactive Effects of an Herbivore-Induced Plant Volatile and Color on an Insect Community in Cranberry
Round 1
Reviewer 1 Report
This paper describes experiments to investigate the effects of colour and odour and the interaction between the two on the attraction of predators, parasitoids and herbivores to traps in cranberry bogs. The paper is a useful contribution in being more comprehensive than most other studies in terms of the range of insect guilds studied and consideration of the interaction between colour and odour. The results are interesting from both academic and applied point of view in demonstrating how different species are affected in very different ways.
The main queries relate to experimental design and statistical analysis
L104. Please justify why were traps of the same colour paired? As the positions of lure and non-lure traps of the same colour were always next to each other and not completely randomised, this will have some implications for statistical analysis (see below).
Was the relative position of the lure and non-lure traps in each pair alternated between weeks, or between replicates? This will also have implications for separating position and presence or absence of lures in statistical analysis.
L135. The authors begin by stating that the analysis was undertaken as a 5 x 2 factorial design, but this is not consistent with L140-142 which gives the full model. Some clarification of the statistical procedures applied is required (see below).
Data analysis
It is unclear how the data analyses have been performed and reported. The authors state that they have applied a repeated measures ANOVA design to the analysis, treating insect counts from each bog as either a replicate (L96) or as blocked random effect (L141), the latter term more commonly associated with mixed effects models.
If a repeated measures ANOVA has been conducted, the authors need to report both the within and between block effects, which will have separate residual degrees of freedom. Here, only the total residual df is given, which if used in calculation of P values from F tests, will lead to serious overestimates of significance.
As these data have elements of both spatial and temporal pseudoreplication (repeated insect counts taken from the same bogs over time, with Predalures not replaced) a repeated measures ANOVA is probably not appropriate. A more effective approach would be to apply one of the mixed modelling packages in R, which can account for multiple random effects.
For example, using the lme4 package:
Model.1<-lmer(Insect counts ~ Color * Predalure + (1|Bog)+(1|Date))
An addition sub-bog random effect may be necessary to account for pairing between traps of the same colour.
Significance of fixed effects terms in the model can be assessed using the anova() command comparing models with and without the given term. Full details are given in the The R Book (Crawley).
This approach may not be possible in combination with MANOVA. Instead, the authors could explore whether total numbers of species or genus from each guild caught differs between treatments, prior to focusing on individual genera.
Tables 1 and 2. Where interactions are found between terms in the model, the significance of the individual terms becomes irrelevant. For example, the interaction between Predalure, Color and Date indicates that the interaction between Predalure and Color is inconsistent between Dates. Treating Date as a random effect (see above) would remove the need for the three-way interaction. The emmeans package is useful for conducting post hoc tests between factor levels in mixed models, taking interactions into account.
Other specific comments
L17,79. Should be “visual clues e.g. color”. There are plenty of visual cues other than colour – shape, contrast, patterns, etc.
L94, 170, 171, 173 etc. . Italicise Latin names
L117. These days it is conventional for experiments on colour to specify colours more accurately rather than just the human perception of red, blue, etc. This makes for a greater understanding of what the insects are perceiving and for more reliable comparison of results from different authors. Ideally reflectance curves should be measured, but other measures such as RGB numbers, Pantone colours are useful. There is also the issue of UV reflectance which can be very important in insect responses and e.g. white substrates often include fluorescent brighteners.
Fig. 2. Are these for colours with or without MeSal? If combined, is this relevant as significant colour X lure interaction in several cases?
Fig. 3, 4. Is it useful to quote differences with 0.05 < P < 0.1, or misleading? In which case, L176 needs rewording as the addition of Predalure to the blue traps does not cause a significant increase in catches (P > 0.05)
Author Response
We have carefully read the comments from reviewer 1 and have addressed all the reviewer's comments and editorial suggestions in the revised manuscript.
Details on how these comments were addressed are presented below (our responses are in bold).
Reviewer 1
The paper is a useful contribution in being more comprehensive than most other studies in terms of the range of insect guilds studied and consideration of the interaction between colour and odour. The results are interesting from both academic and applied point of view in demonstrating how different species are affected in very different ways.
We thank Reviewer 1 for the positive comments.
L104. Please justify why were traps of the same colour paired? As the positions of lure and non-lure traps of the same colour were always next to each other and not completely randomised, this will have some implications for statistical analysis (see below).
We paired them because we wanted to avoid potential positional biases. We have indicated this in the text. Line 94.
Was the relative position of the lure and non-lure traps in each pair alternated between weeks, or between replicates? This will also have implications for separating position and presence or absence of lures in statistical analysis.
No, the position of the lure and non-lure traps in each pair was, however, not alternated between weeks but their position were randomly assigned among replicates. We have now indicated this in the text. Lines 101-103.
L135. The authors begin by stating that the analysis was undertaken as a 5 x 2 factorial design, but this is not consistent with L140-142 which gives the full model. Some clarification of the statistical procedures applied is required (see below).
We have followed the advice of Reviewer 1 and re-analyzed the data. See below. This has made the story simpler and clarified the analysis.
It is unclear how the data analyses have been performed and reported. The authors state that they have applied a repeated measures ANOVA design to the analysis, treating insect counts from each bog as either a replicate (L96) or as blocked random effect (L141), the latter term more commonly associated with mixed effects models.
We have followed the advice of Reviewer 1 and re-analyzed the data. See below.
If a repeated measures ANOVA has been conducted, the authors need to report both the within and between block effects, which will have separate residual degrees of freedom. Here, only the total residual df is given, which if used in calculation of P values from F tests, will lead to serious overestimates of significance.
As indicated, we have followed the advice of Reviewer 1 and re-analyzed the data. We are not using repeated measures ANOVA in the new analysis. See below.
As these data have elements of both spatial and temporal pseudoreplication (repeated insect counts taken from the same bogs over time, with Predalures not replaced) a repeated measures ANOVA is probably not appropriate. A more effective approach would be to apply one of the mixed modelling packages in R, which can account for multiple random effects.
For example, using the lme4 package:
Model.1<-lmer(Insect counts ~ Color * Predalure + (1|Bog)+(1|Date))
An addition sub-bog random effect may be necessary to account for pairing between traps of the same colour. Significance of fixed effects terms in the model can be assessed using the anova() command comparing models with and without the given term. Full details are given in the The R Book (Crawley).
We have followed the reviewer’s comment and re-analyzed our data using the model suggested by the reviewer (lines 141-143). We were familiar with this model since we had used it in past data analyses. The general conclusions have remained the same as previously, but the story and analysis are now simpler. We thank the reviewer for the statistical advice.
This approach may not be possible in combination with MANOVA. Instead, the authors could explore whether total numbers of species or genus from each guild caught differs between treatments, prior to focusing on individual genera.
As the reviewer indicates, we were unable to use the model with MANOVA. For the MANOVA, we used the total numbers of each guild across all weeks.
Tables 1 and 2. Where interactions are found between terms in the model, the significance of the individual terms becomes irrelevant. For example, the interaction between Predalure, Color and Date indicates that the interaction between Predalure and Color is inconsistent between Dates. Treating Date as a random effect (see above) would remove the need for the three-way interaction. The emmeans package is useful for conducting post hoc tests between factor levels in mixed models, taking interactions into account.
As indicated by the reviewer, after using the new model, we have eliminated the three-way interactions. Thus, the Tables and Figures have been changed accordingly. Although the general conclusions have not changed with the new analysis, the story simpler.
Other specific comments
L17,79. Should be “visual clues e.g. color”. There are plenty of visual cues other than colour – shape, contrast, patterns, etc.
We have added “e.g. color” as suggested by the reviewer. Lines 15, 76.
L94, 170, 171, 173 etc. . Italicise Latin names
We are not sure why the species names did not show in italics when in the original manuscript all scientific names were italicized. Something happened during the submission process. We will make sure all the scientific names are italicized.
L117. These days it is conventional for experiments on colour to specify colours more accurately rather than just the human perception of red, blue, etc. This makes for a greater understanding of what the insects are perceiving and for more reliable comparison of results from different authors. Ideally reflectance curves should be measured, but other measures such as RGB numbers, Pantone colours are useful. There is also the issue of UV reflectance which can be very important in insect responses and e.g. white substrates often include fluorescent brighteners.
We agree with the reviewer. We did not include the reflectance curves for the colored traps used because these data have been published previously. We used the same colored traps used in a previous study (Silva et al. 2018). We have added a sentence in the methods that says: “Reflectance data for these colored traps are reported in Silva et al. [42].” Line 114.
Fig. 2. Are these for colours with or without MeSal? If combined, is this relevant as significant colour X lure interaction in several cases?
We agree with the reviewer that, if the color-by-lure interaction is significant, no need to include it in this Figure. This is the case for hoverflies; thus, we have removed them from the figure. We have also changed the order of the figures in the revised manuscript. The now Figure 3 includes only the insect families where color (and not MeSA or their interaction) was significant.
Fig. 3, 4. Is it useful to quote differences with 0.05 < P < 0.1, or misleading? In which case, L176 needs rewording as the addition of Predalure to the blue traps does not cause a significant increase in catches (P > 0.05)
We have removed the non-significant p values in the now Figure 2 (Figure 4 has been removed because of the new statistical analysis, see above). We would like to mention in the text the fact that blue traps baited with PredaLure caught 2x more hoverflies than unbaited traps even when this difference was marginal. We believe this is important to note and have added a sentence: “Although blue traps baited with PredaLure captured 2 times more T. marginatus than unbaited blue traps, this difference was only marginally significant (P = 0.08; Figure 2).” Lines 172-174.
Reviewer 2 Report
This is an interesting time limited study that draws attention towards the use of selective colors, chemical cues and sticky traps in the cranberries field to attract or repel various groups of insects. My suggestions and comments on this communication paper are provided below:
L16: Delete remains
L18: After traps insert ‘in field conditions’
L22 After thrips insert species scientific name
L23: Delete respectively
L27: Delete s in implications
L42: Delete known to be
L46: Use parenthesis for (plant-herbivore-natural enemy)
L50: Delete ‘feeding by’ and insert feeding after herbivores
L56: Insert species scientific names after hoverflies
L61: Delete s after numbers
L69: Change hoverflies to hoverfly
L111: Insert Direction of North in Figure 1, with an arrow pointing up
L138: After MANOVA insert (Multivariate Analysis of Variance)
L16: 42 After ANOVA insert (Analysis of Variance)
L150: Insert d after indicate
L168: Were these multiples species, if so, insert
L170, 171, 173, 176, 178, 179, 181, 195, 198: Use Italics for the species underlined
L182: Delete such after with
L197-198: Delete one of the sampling date, namely, and , respectively
L210: After (Figure 2I) insert . and delete and.
L211: Change click to Click
L227: Delete we before showed and insert ‘this study’
L243: Delete will before optimize
L248: Delete the colors before white and insert colors after blue
L250: Delete we before showed and insert our results
L254: Delete attractive before response
Conclusion Section: LN312-323:
Overall the conclusion needs certain improvement with respect to insect community attracted to specific traps. Let's say what predators, parasitoids, and pollinators were attracted and how MeSA based tools can help to manage pests (species names) in the cranberries. Use of stick traps (different color) and MeSA attract specific herbivores (pests) and beneficial species. If more beneficial species are caught in sticky traps than pests. How these practices will help to
conserve beneficial species in general. Were there any chances of the flying insects moved from other fields. If so, a sentence can help the readers.
Comments for author File: Comments.pdf
Author Response
We have carefully read the comments from reviewer 2 and have addressed all the comments and editorial suggestions in the revised manuscript.
Details on how these comments were addressed are presented below (our responses are in bold).
Reviewer 2
This is an interesting time limited study that draws attention towards the use of selective colors, chemical cues and sticky traps in the cranberries field to attract or repel various groups of insects.
We thank the reviewer for the positive remarks and editorial comments.
L16: Delete remains
We have changed to “it is unknown”. Line 15.
L18: After traps insert ‘in field conditions’
Because of word limits we are indicating this in the next sentence. Line 17.
L22 After thrips insert species scientific name
We did not identify the thrips to species, so we are not able to include this. We discuss the thrips species we have found in cranberries in past studies which likely are the same caught in our traps. Lines 276-277.
L23: Delete respectively
This has been deleted
L27: Delete s in implications
We would like to keep “implications” as we are talking about “findings” which is plural.
L42: Delete known to be
We have deleted this.
L46: Use parenthesis for (plant-herbivore-natural enemy)
We have added parenthesis as suggested by the reviewer. Line 44.
L50: Delete ‘feeding by’ and insert feeding after herbivores
We have changed to “by herbivory from insects”. Line 48.
L56: Insert species scientific names after hoverflies
The species names of hoverflies are not reported in this study, so we would like to keep as is.
L61: Delete s after numbers
We would like to keep as “numbers”, plural.
L69: Change hoverflies to hoverfly
We would like to keep as hoverflies for consistency since other predator groups are mentioned in plural.
L111: Insert Direction of North in Figure 1, with an arrow pointing up
As suggested by the reviewer, we have added a compass in Figure 1.
L138: After MANOVA insert (Multivariate Analysis of Variance)
We have spelled out this as suggested. Line 137.
L16: 42 After ANOVA insert (Analysis of Variance)
We have spelled out this as suggested. Line 141.
L150: Insert d after indicate
We have deleted this sentence in the revised manuscript.
L168: Were these multiples species, if so, insert
This section has been modified slightly due to the new data analyses, suggested by Reviewer 1. We have mentioned species identity when available.
L170, 171, 173, 176, 178, 179, 181, 195, 198: Use Italics for the species underlined
As indicated in our response to Reviewer 1, we will make sure all the scientific names are italicized.
L182: Delete such after with
This has been deleted.
L197-198: Delete one of the sampling date, namely, and , respectively
This has been deleted in the new version of the manuscript.
L210: After (Figure 2I) insert . and delete and.
We have changed this to “, while click”. Line 188. We would like to keep this in a single sentence.
L211: Change click to Click
Please see previous comment.
L227: Delete we before showed and insert ‘this study’
As suggested, we have changed “we” for “this study”. Line 213.
L243: Delete will before optimize
We have deleted “will”. Line 229.
L248: Delete the colors before white and insert colors after blue
We would like to keep as is because we think it reads a bit better.
L250: Delete we before showed and insert our results
We have changed as suggested by the reviewer. Line 235.
L254: Delete attractive before response
We have deleted as suggested by the reviewer.
Conclusion Section: LN312-323:
Overall the conclusion needs certain improvement with respect to insect community attracted to specific traps. Let's say what predators, parasitoids, and pollinators were attracted and how MeSA based tools can help to manage pests (species names) in the cranberries. Use of stick traps (different color) and MeSA attract specific herbivores (pests) and beneficial species. If more beneficial species are caught in sticky traps than pests. How these practices will help to conserve beneficial species in general. Were there any chances of the flying insects moved from other fields. If so, a sentence can help the readers.
We thank the reviewer’s comment and have changed the “Conclusions” section to improve readability. We now start with the insects affected by MeSA-baited colored traps and the potential for using these traps to monitor these insects. We then discuss how these traps could be used to help conserve beneficial insects. We don’t know the range of attraction for these traps so, at this point, we cannot make any conclusions on the chances of attracting insects from other fields. To address this, we end the conclusions by stating “To implement MeSA-baited colored traps to monitor hoverfly and thrips populations in cranberries, further studies are needed to estimate their range of attraction.”
Round 2
Reviewer 1 Report
The authors have generally done a good job responding to the Reviewers' comments.
Minor comments.
The authors should update Tables and Figures to make clear where they have used the mixed-model analyses
L94. not totally sure how pairing the traps of the same colour "avoids positional bias". If anything I would have thought it increased it relative to a completely randomised layout?
L174. given that the authors specify P<0.05 as their significance level, P = 0.08 is not "marginally significant". It could be "marginally insignificant" or, best, "not significant"
The new reference 42 is wrongly quoted
Journal of Insect Science, Volume 18, Issue 2, March 2018, 19, https://doi.org/10.1093/jisesa/iey005
Author Response
The authors have generally done a good job responding to the Reviewers' comments.
We thank the reviewer for all the helpful comments
Minor comments.
The authors should update Tables and Figures to make clear where they have used the mixed-model analyses
We have added this information to the Tables 2-4.
L94. not totally sure how pairing the traps of the same colour "avoids positional bias". If anything I would have thought it increased it relative to a completely randomised layout?
In our experimental design, we did not want to bias the position of traps in relation to their proximity to the forest habitat. This is why we decided to pair them. We have change the text to indicate this.
L174. given that the authors specify P<0.05 as their significance level, P = 0.08 is not "marginally significant". It could be "marginally insignificant" or, best, "not significant"
We have changed to read “only marginal and statistically insignificant”
The new reference 42 is wrongly quoted
Journal of Insect Science, Volume 18, Issue 2, March 2018, 19, https://doi.org/10.1093/jisesa/iey005
We have changed the reference