Comparison of Two Sampling Methods to Estimate the Abundance of Lucanus cervus with Application of n-Mixture Models

Round 1

Reviewer 1 Report

This paper presents a comparison of abundance and population estimates of beetles observed using two sampling methods. I think that the paper provides good perspectives on how to effectively sample populations of species of concern. My overall critique of this paper is that the authors leave too much information unexplained or undefined, forcing the reader to piece together significant methodologies, results, and discussion. This issue is not insurmountable. It only requires a more careful presentation of methods and results, so that the discussion associated with them makes more sense and is more intuitive to the reader. Some of the things the authors leave the reader to assume are:

• Definition of ‘costs’ related to the specific study
• RRC (line 93) is the modeling approach used (line 106)
• ‘complex mathematical modelling’ (lines 94-95) is the same as Bayesian framework (line 106 and later)
• Comparison of ‘two sampling methods to derive accurate abundance estimates while accounting for imperfect detection…..’ (lines 103-104) is the same as comparing costs and benefits (title and discussion)
• Location of study site (no coordinates)
• Sources of variables for detection probability for N-mixture models (listed in Table 1)
• Logic behind analyzing transects individually, when they seem at first read that they are replicates of a single sampling method
• Ranges/means of significant variables (Tables 1 and 2)

I have included specific comments and suggestions in the body of the text using the pdf Comment tool. Please address all comments and think about how to relate all parts of the paper to one another for clarity.

Comments for author File: Comments.pdf

Author Response

Reviewer 1: My overall critique of this paper is that the authors leave too much information unexplained or undefined, forcing the reader to piece together significant methodologies, results, and discussion. This issue is not insurmountable. It only requires a more careful presentation of methods and results, so that the discussion associated with them makes more sense and is more intuitive to the reader. Some of the things the authors leave the reader to assume are:

Authors: We replied to all the comments rise by this Reviwer below and in the file in attachment.

 

Reviewer 1: Definition of ‘costs’ related to the specific study.

Authors: We agree, and also following the suggestion of the Rev2 we changed the title in “Comparison of ‘two sampling methods….” Therefore we discuss about costs and benefit only in the discussion. In this section, a specific paragraph clearly explains what we mean by “cost and benefit”.

 

Reviewer 1: RRC (line 93) is the modeling approach used (line 106).

Authors: We specified it also in the section ‘Modelling procedures’ of the ‘Materials & Methods’, accordingly.

 

Reviewer 1: ‘complex mathematical modelling’ (lines 94-95) is the same as Bayesian framework (line 106 and later).

Authors: Yes. We specified “Bayesian framework” in the text, accordingly.

 

Reviewer 1: Comparison of ‘two sampling methods to derive accurate abundance estimates while accounting for imperfect detection…..’ (lines 103-104) is the same as comparing costs and benefits (title and discussion).

Authors: Yes, we agree and, also following the suggestion of the Rev2, we changed the title in “Comparison of two sampling methods…”.

 

Reviewer 1: Location of study site (no coordinates).

Authors: We specified the coordinates, accordingly.

 

Reviewer 1: Sources of variables for detection probability for N-mixture models (listed in Table 1).

Authors: We specified the source of variables in the text, accordingly.

 

Reviewer 1: Line 149. Why were 13 trees selected and not 12 or 15 or 23? As presented, this number seems arbitrary (no context or explanation). Is this the result of a statistical power calculation? Or, were there natural micro-physiographic sections or partitions within the study site that made this number of trees intuitive? More explanation is needed to inform the reader why this number of trees is adequate or appropriate to compare with sampling conducted across 1000m of ecotone (in other method).

Authors: We specified in the text the criteria for choosing those 13 trees.

 Reviewer 1: Six of your 13 survey trees are within two pixels of your transect routes. What of concerns of sampling being spatially biased? What are the implications of having about half of your survey points for one sampling method in such close proximity to points associated with the other sampling method?

Authors: We understand the point raised by this reviewer but we do not see any reasons for which having trees partially matching the pixel in which transects occurred could bias our analyses. On the other side we believe that having pixels in which trees and transects matched will strengthen our comparison.

 

Reviewer 1: Table 1. Since you have units, why not present the range of values for each variable? Although the VIF is important statistically, the ranges of values are more important for readers who may want to conduct similar surveys under similar conditions.

Authors: Done, we displayed in Table 1 the ranges of all covariates, accordingly.

 

Reviewer 1: Line 171 and 175. I think this can be cited in the 'References' and not included long-form in the text.

Authors: According to the journal guidelines we prefer living the links here.

 

Reviewer 1: Line 179. Add sentences stating why individual transects were analyzed and how that ties into the objectives of this paper.

Authors: Done.

 

Reviewer 1: Logic behind analyzing transects individually, when they seem at first read that they are replicates of a single sampling method. Authors: Yes, they are replicates of the same method. However, according to the standard protocol for the monitoring of L. cervus (Bardiani et al., 2017), the number of transects per area can vary from 1 to 4 depending on the size of the area to be investigated. Therefore, in order to evaluate the improving of the sampling and the estimate accuracy from one to two transects, it was necessary to consider them individually. Moreover each transect, cover a small and partial area of the forest while trees are more homogeneously distributed and cover the main part of the forest. 

Reviewer 1: Ranges/means of significant variables (Tables 1 and 2).

Authors: Done, we displayed in Table 1 the full name of all covariates, their means and ranges accordingly.

 

Reviewer 1: What is the purpose of looking at each transect separately? No corresponding subset of tree counts was conducted to see population estimates based on a subset of individual tree counts. This study was designed with two sampling methods with replicates of samples within each method. Therefore, why would you isolate your sampling replicates for either method, let alone conduct analysis of each individual replicate of only one method and not the other? To me, this analysis of individual transects is not intuitive or needed. If this analysis is retained in the Results (in text and Figure 2) and Discussion, then a justification and/or explanation for it needs to be included in the analysis section of the Methods, so that the reader understands the logic of this aspect of analysis.

Authors: While on the one hand the guidelines for the monitoring of the stag beetle consider a minimum number of transects equal to 1 (therefore a single replica according to the Reviewer 1 point of view) sufficient to obtain a sample of individuals representative of the population, a single tree will never be sufficient to obtain a representative sample of the population. Therefore on the basis of this consideration, we cannot consider a single tree as a single transect. Comparison is absolutely not possible. Instead, it is possible to consider a certain number of trees equivalent to a single transect (see Vrezet at al., 2012, the only research before us that tested the tree trunk surveys method (nocturnal), considered 10 trees as a minimum monitoring unit. In our case, we started with the idea of ​​looking for all the trees suitable for the species and in the end we identified 13 trees. We therefore assumed the number 13 trees as the minimum unit of monitoring (assuming that there were no other trees that the species could occupy to feed). At that point, our goal was to verify whether, with the walk transects method (considered the most effective in the literature) it would be possible to obtain a population estimate comparable to that obtained with the tree trunk survey method and, above all, to evaluate whether it was necessary to implement more than one transect in order to achieve this goal. (See sentences we added in the text “Modelling procedures” section).

 

Reviewer 1: Line 223. The upper and lower bounds of this Cr.I. are the sum of the upper and lower bounds (respectively) of the individual transects. Is that correct? I am unfamiliar with how Cr.I. works, but it doesn't seem that the intervals should be the arithmetic sum of the upper and lower bounds of component samples. If you calculated individual Cr.I. for each of the survey trees would the lower bound sum to 120 and the upper bound sum to 199?

Authors: we don’t understand the comment here as actually these are the values we provided in the manuscript as lower and upper bounder of the abundance estimates.

 

Reviewer 1: To me this seems counter intuitive, because as you add sample units (trees or transects) you would expect your estimates to get better/more accurate. However, if Cr.I. bounds increase as sampling increases the estimate would be worse (greater range between upper and lower bounds and larger bounds seem to reflect a less accurate estimate).

Authors: We understand the point raised here but this is what mathematically happened when you include (not sequentially add) trees/portion of transects with heterogeneous number of individuals. Actually, landscape characteristics are not homogeneous in our study are and this is the reason why we included landscape predictors to model species abundance.

 

Reviewer 1: This figure needs to be placed before Table 1 in the manuscript. Table 1 is not mentioned until page 4, so there is room to move the figure to follow the first mention of Figure 1 on page 3.

Authors: we deleted Figure 2 as suggested also by the Reviewer 2.

 

Reviewer 1: line 271 For this study, were two operators used for the transects (one per transect)? In the methods there was no distinction that one operator was used to conduct tree counts and two were used to conduct transects. I do not see the need for two transect operators.

Authors: We specified in the text (“Materials and Methods” section) that there was one operator per transect.

 

Reviewer 1: paragraph “cost and benefits in the “discussion” section. You need to define your costs, so that the benefits are more obvious. For instance, I would view sampling time as a cost, so the requirement of two transects (with either one or two operators) at 1-hour total is better compared to 2-3 hours surveying individual trees. If costs are defined then the reader is more likely to follow your logic in promoting one method over the other.

Authors: We agree that the terms cost and benefits can be misleading, therefore we changed the paragraph title in Pros & Cons. (as well as the manuscript title).

 

Reviewer 1: I have included specific comments and suggestions in the body of the text using the pdf Comment tool. Please address all comments and think about how to relate all parts of the paper to one another for clarity.

Authors: we carefully address all comments provided by this Reviewer in the body of our manuscript.

Author Response File: Author Response.docx

Reviewer 2 Report

In this work the authors compare two methods to estimate the abundance of Lucanus cervus. The use of N-mixture models further allowed to determine significant variables explaining the abundance of the target species.

The work is relevant to improve monitoring methods of this species which is under a EU conservation directive. The text is in general well written, still a few typos need to be corrected. Also frequently sentences are too long.

The title starts with Cost and Benefits suggesting an economic analysis which is not the case: I suggest maybe to replace Cost and Benefits by Comparison

The abstract in my opinion does not highlight the relevance of the work.

The abstract could start focusing the relevance of this particular species and its conservation status. Then explain the need for accurate and practical monitoring methods.  Then say what was done on this work. The text on L17-L20 could be removed.  Text in bracket in L28 should be removed as well. The same for the text in L30. On the contrary, a sentence explaining the environment variables explaining the abundance of Lucanus cervus (distance to forest, tree cover, etc.) is lacking in the abstract, which I believe would be a relevant information.    

As a reader I feel that Methods and Results sections in some cases need more detailed information

L136 – L145 - All this part of the text should go to Introduction,

L149-L151 – Up to which height from the ground were the observations carried on? How tall were the trees?  At what time of the day were diurnal tree survey carried on? How long did you stay observing each tree?

L149-L151 - What was the total area covered by the 13 oak trees? This issue is particular relevant as you aim to estimate the abundance of L. cervus, however it depends on the area covered by your survey!

L197-198 – Does these data refers to the accumulated observations over time (6 observations) on the same pixel or a mean per each observation? Not clear to me.

L214 – According to Table 2 should be increase with increasing distance to grassland (not decreasing!). An increase with increasing distance is also what would make more sense!

L222-L226- This rationale makes no logic to mean! From Figure 1 it is clear that one transect covers about half, maybe less than half, of the total area sampled when considering the 13 oak trees! So, it is not so surprise that the abundance estimated by one transect alone would be half of the tree surveys!

L241 – Delete the word “the”

L241-244 - Break this long sentence in two!

L269-L270 – You are being too simplistic by generalizing these results! Maybe other factors such as host tree density (as you demonstrated), host tree species, tree height, etc. will affect L. cervus detectability and therefore the number of trees needed to survey.

L272-275 – See my comment above, you cannot compare what is not comparable! If one transect covers only about half of the area covered by the 13 oaks, it seems logic that you obtain only half of the estimates!

L285 – correct the word - Lucanus

Figure 1 – The scale in the bottom must be wrong! You say that each transect is about 500m and the area is about 20ha, but according to the scale in the figure each transect should have more than 3 Km and total area should be more than 20km²!

Figure 2 is not needed as this information is on the text. I suggest to delete it

L384- Simplify, start sentence with “Although …”

L385 – Delete the word “the”

Author Response

Response to Reviewer 2

 

Reviewer 2: The title starts with Cost and Benefits suggesting an economic analysis which is not the case: I suggest maybe to replace Cost and Benefits by Comparison

Authors: Done.

 

Reviewer 2: The abstract could start focusing the relevance of this particular species and its conservation status.

Authors: Done.

 

Reviewer 2: Then explain the need for accurate and practical monitoring methods.  

Authors: Done.

 

Reviewer 2: Then say what was done on this work.

Authors: Done.

 

Reviewer 2: The text on L17-L20 could be removed. Text in bracket in L28 should be removed as well. The same for the text in L30.

Authors: Here we disagree and left the text as it was because we believe is important to provide useful information to the readers.

 

Reviewer 2: On the contrary, a sentence explaining the environment variables explaining the abundance of Lucanus cervus (distance to forest, tree cover, etc.) is lacking in the abstract, which I believe would be a relevant information.    

Authors: Done.

 

Reviewer 2: As a reader I feel that Methods and Results sections in some cases need more detailed information.

Authors: Information are improved following also Rev 1 suggestions.

 

Reviewer 2: L136 – L145 - All this part of the text should go to Introduction,

Authors: I’m not agree, I specifically named this paragraph “Study species and data collection” to be able to describe in more detail the species.

 

Reviewer 2: L149-L151 – Up to which height from the ground were the observations carried on? How tall were the trees?  At what time of the day were diurnal tree survey carried on? How long did you stay observing each tree?

Authors: We specified all these information in the main body of our manuscript (“Materials and Methods”) and also in Table 1.

 

Reviewer 2: L149-L151 - What was the total area covered by the 13 oak trees? This issue is particular relevant as you aim to estimate the abundance of L. cervus, however it depends on the area covered by your survey!

Authors: We understand the concern raised by the Reviewer here but we stress that the aim of our study is to compare two sampling methods to estimate abundance of L. cervus not to estimate the optimal number of transects and trees to estimate species abundance. Thus, back to the question of the Reviewer, we directly refer to the number of pixels considered in our study for each sampling methods. Actually, the total number of pixels covered by the transects is respectively 72 and 49, while those of the trees is 13 and thus we believe that even if one transect covers approximately about half of the area covered by the 13 trees, it is not obvious that we obtained only half of the estimates.

 

Reviewer 2: L197-198 – Does these data refers to the accumulated observations over time (6 observations) on the same pixel or a mean per each observation? Not clear to me.

Authors: This is the accumulated number of observations over time (6 observations).

 

Reviewer 2: L214 – According to Table 2 should be increase with increasing distance to grassland (not decreasing!). An increase with increasing distance is also what would make more sense!

Author: We corrected it accordingly.

 

Reviewer 2: L222-L226- This rationale makes no logic to mean! From Figure 1 it is clear that one transect covers about half, maybe less than half, of the total area sampled when considering the 13 oak trees! So, it is not so surprise that the abundance estimated by one transect alone would be half of the tree surveys!

Author: We do not understand the concern of the Reviewer here. In the text we already specified that “When deriving the total population size for both the methods we found relatively similar results, i.e. 157.21 (95% Cr.I. 120-199) for tree surveys and 159.71 (95% Cr.I. 71-434) for walk transects. However, when we included in the models only one transect at a time, the total population size estimated for single transect dramatically decreased compared to that estimated for tree surveys, i.e. 88.66 (95% Cr.I. 37-257) for T1 and 71.5 (95% Cr.I. 34-177)”, this logic make sense to us.

           

Reviewer 2: L241 – Delete the word “the”

Author: Done.

 

Reviewer 2: L241-244 - Break this long sentence in two!

Author: Done.

 

Reviewer 2: L269-L270 – You are being too simplistic by generalizing these results! Maybe other factors such as host tree density (as you demonstrated), host tree species, tree height, etc. will affect L. cervus detectability and therefore the number of trees needed to survey.

Author: We understand the point raised by this Reviewer and thus we provided more details on the trees considered in this new version of our manuscript. We agree that other factors may affect L. cervus detectability on trees but we retained only the predictors considered to match those considered for transect sampling for comparison.

 

Reviewer 2: L272-275 – See my comment above, you cannot compare what is not comparable! If one transect covers only about half of the area covered by the 13 oaks, it seems logic that you obtain only half of the estimates!

Author: See our answer above. The total number of pixels covered by the transects is respectively 72 and 49, much higher than those of the trees (N=13) and thus we believe that even if one transect covers approximately about half of the area covered by the 13 trees, it is not obvious that we obtained only half of the estimates.

 

Reviewer 2: L285 – correct the word – Lucanus

Author: Done.

 

Reviewer 2: Figure 1 – The scale in the bottom must be wrong! You say that each transect is about 500m and the area is about 20ha, but according to the scale in the figure each transect should have more than 3 Km and total area should be more than 20km²!

Authors: Yes, this is indeed a leftover and now we provided a new figure 1 with correct scale values.

 

Reviewer 2: Figure 2 is not needed as this information is on the text. I suggest to delete it

Authors: We deleted figure 2, accordingly.

 

Reviewer 2: L384- Simplify, start sentence with “Although …”

Authors: Done.

 

Reviewer 2: L385 – Delete the word “the”

Authors: Done.