Who and Where Are the Observers behind Biodiversity Citizen Science Data? Effect of Landscape Naturalness on the Spatial Distribution of French Birdwatching Records
, Sébastien Caillault 2, Joséphine Pithon 1, Guillaume Pain 1, Hervé Daniel 1, Benoit Marchadour 4 and Véronique Beaujouan 1,*
Rodrigues Pedro
Round 1
Reviewer 1 Report
In this paper, the authors analyze the effect of CS observer profile on the spatial distribution of recordings from a French bird CS program. The authors hypothesized that observer profile biased the spatial distribution of recordings and that this bias could be explained by the naturalness of the landscape. They developed observer profiles based on analysis of the temporal and spatial distributions of their recordings as well as the content of the recordings. They mapped a gradient of naturalness at regional and local scales. They then defined four types of observers: gardeners, beginners, naturalists, and experts. They found that recording intensity could be related to naturalness at the regional level (the most visited areas were those with lower average naturalness and high accessibility due to a well-developed road infrastructure). At the local level, they found that experts and naturalists recorded in areas with a higher naturalness index than home gardeners and beginners.
The article is overall clear and well-constructed. The starting hypotheses and results are interesting even if some results are expected (for example the fact that naturalists tend to explore more natural areas). Nevertheless, this does not detract from the quality of the article, which has the merit of scientifically proving what is often suggested. The discussion is also well constructed and sufficiently supported in terms of bibliography. The main point to improve the manuscript is the material and the method. Indeed, some elements are missing to clearly understand the choices made (whether in terms of analysis, choice of scale...) and additional information on the monitoring protocol is missing.
Find here more specific comments.
Title.
This is a personal feeling, but I think the title is very long. We are often faced with the dilemma between a short and unclear title and a long but complete title. Perhaps there is a middle ground to be found here. But this is just a personal opinion and if a choice has to be made, we might as well leave a long but complete title.
The introduction is well constructed and clear. Here are some minor suggestions.
L48. I disagree, many mass CS experiments also provide quantitative data. These approaches are not limited to just occurrence data.
L 51-65. The authors speak directly of bias. But at no point are the types of bias described. CS generates many other biases, so why only mention these? Before giving examples, it would be useful to better present the different categories of biases in order to situate those mentioned in this paper among those well known in CS.
L100-104. It would be useful to associate specific assumptions with bird databases (showing the specificity of the latter). Indeed, these kind of data refer to songs (mainly), whereas in many other CS experiments, they are direct observations. Therefore, is it more difficult or easier to recognize birds based on songs? Do studies of song listening show a difference between experts and novices? A little more specific work on this type of data would be a safe bet for introducing hypotheses, in my opinion.
Methods.
General comments.
Some information about the CS protocol itself is missing to understand what type of data is being generated. Also, descriptions of the 22 indicators would be helpful in the appendix. Another possible method (part 2.4.3) would have been to use the Multidimensional Scaling (MDS) which allows to pass from a proximity matrix (similarity or dissimilarity) between a series of N objects to the coordinates of these same objects in a p-dimensional space. For example, with MDS, it is possible to reconstruct very precisely the position of samples on a map from the kilometer distances (the dissimilarity is then a Euclidean distance) between cities, natural areas...
More details would be useful to understand how these data are collected (are there direct observations, songs...). Also, what are the criteria to participate (are there any prerequisites...). It was noted in section 2.2 that the objectives, methods, and data collection have changed. Does this have an impact on data collection and especially on the users who are doing the monitoring?
I do not understand why the authors did not perform a spatial aggregation analysis between the sample points and the NI? By using Moran index or something else? Another possible method would have been to use the Multidimensional Scaling (MDS) which allows to pass from a proximity matrix (similarity or dissimilarity) between a series of N objects to the coordinates of these same objects in a p-dimensional space. For example, with MDS, it is possible to reconstruct very precisely the position of samples on a map from the kilometer distances (the dissimilarity is then a Euclidean distance) between cities, natural areas...
Some specific comments.
L113; "as far as is possible". What does it mean?
L120. Replace "verified" by "checked"
L131. Details of all 22 parameters (not just the 11) should be added to Appendix A1.
Table 2. Please add SD to mean because we are talking about variation in the title, but only the means are presented.
L140-141. Can you specify the reasons because I don't understand these two choices.-.
L143-146: To determine the number of groups without a priori, it is also possible to use the silhouette width method S (Maechler et al. 2021).
L150. A bracket is missing
Figure 1. (i) Remove the legend from the OP? While I appreciate the effort of the figure, it is still not very clear for me. Maybe, you can insert the figure directly in part 2.4.1?
L230. Why a 2*2km grid?
L235. You need to justify why 5 classes were selected.
L236-238: If we use transformed data and it still does not allow us to have the conditions to perform parametric tests, then why keep the transformed data?
L240. Why did you use Bonferroni's correction (which is particularly criticized) when you use Dunn's already corrected posthoc test?
L250. Again, you must justify this value. Why 400m²?
Results.
The results are well written and clear.
Some minor suggestions.
Table 2: NR Total number of what?
Fig 1 and 2. Please add a location map, not everyone knows French geography
Fig. 5. Not sure that the zoom is really useful.
Discussion.
Why use personal nomenclatures to determine classes of participants in this study. Indeed, in the discussion, it is mentioned that several other studies already use terms to qualify participants (super user, dabbler...). This can be confusing. If this is justified by the fact that the quality metrics were used in this study compared to others, then this should be clearly stated at some point to justify the use of new terms.
L441. Reservation?
Author Response
Please see the attachment with responses to the four reviewers' comments.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Dear Editors and Authors,
The manuscript adds another piece to the puzzle in understanding the influence of volunteer behavior and their choice of landscape on the census outcome itself. It builds on a solid foundation of previous work by other authors in this area of research. I ask, which of these approaches and tools are feasible to incorporate into routine censuses, bird counts? And isn't the random selection of quadrats (possibly after prior stratification) already practiced in some census programs (for line or point counts) the answer to the authors' challenge at the end of the paper? After all, this is intended to reduce bias in the potential impact of volunteer landscape selection on the results, isn´t it?
From minor issues:
L47-49: the sentence seems to be incomplete
2.3.1. and 2.3.2: These two factors represent the same perspective at first glance. At least the similarity of the first sentences in both paragraphs may be confusing. I recommend modifying the text slightly.
Dendrogram Figure A3 Appendix A is not readable; it extends beyond the edges of the page.
Author Response
Please see the attachment with the responses to the four reviewers' comments.
Author Response File: Author Response.pdf
Reviewer 3 Report
The manuscript " Who and where are the observers behind big citizen science biodiversity datasets? Investigating the effect of landscape naturalness on the spatial distribution of French birdwatching records" by Guetté et al., present an analysis of the spatial bias in an opportunistic data set on bird counts produced by citizen science in the administrative region of Pays-de-la-Loire (Western France).
In relation to the majority of studies concerning citizen science, this study linked the behavior and experience of the observers with the spatial distribution of recordings, finding interesting results that can be used in similar data in other places in order to better plan where observers with different levels of experience should be more focused.
The methodology used in this study was adequate.
The results are easy to understand. They found that recording intensity could be related to the proximity to populated areas and road infrastructures, meaning less naturalness, and that more experienced observers recorded in more natural areas than the less experienced ones.
The discussion explains the results and reveals when this type of studies should be applied.
I only suggest the authors develop better the figure 1, or at least make it more visible as it is difficult to see the different intensity patterns on the maps; probably on the online paper it will be possible to see a setter image, but on PDF could not be the best. The same for figures 2 and 3. The figure A3 (Appendix), the dendrogram, was impossible to see well, at least in my PDF.
Author Response
Please see the attachment with the responses to the four reviewers' comments.
Author Response File: Author Response.pdf
Reviewer 4 Report
This study gives significant insights into the value of data gathered through citizen science projects. The authors cleverly combine datasets on the profiles of citizen science practitioners with spatial data to tease out these relationships. The analysis is elegant, and the results are presented in informative tables and figures. The complexities of the analysis are fully revealed in useful appendices. The manuscript is well-written, but some minor errors can be addressed as follows.
Line 49: spatially, temporally and taxonomically what? Missing noun, perhaps constraints?
Line 83: For example, a
Line 158: already cited
Line 257: landscapes’
Line 277: analyses and graphs
Line 289: Is precision to two decimal places justified in this and further results? It would be more readable if rounded out to fewer significant figures.
Line 306: (20.69%)
Line 341: Mean NI
Line 346: precision is this table is inconsistent with one or two decimal places – see previous comment above.
Line 314: than that for the other observer profiles
Line 375: 400 m2
Line 380: (Figure F1) – see also lines 393 and 403
Line 493: km2
Line 495-6: Citations inconsistent with journal style
Line 498: type areas
Line 535: Figure A3 did not render correctly in my copy of MS?
Author Response
Please see the attachment with the responses to the four reviewers' comments.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The authors made a serious effort to address the majority of my comments. They also re-analyzed the data taking into account some of the comments made, which is very welcome. In addition, these new analytical analyses confirm their first results, which is a guarantee of the robustness of the results. As I stated in my first review, this paper is of quality and the conclusions are robust. In particular, the work on Figure 1 improves overall understanding. As it stands, I see no reason not to accept the manuscript.
Reviewer 2 Report
Dear Authors,
thank you for responding to my suggestions. After expanding and clarifying some paragraphs, the manuscript now looks coherent. I have no further comments.
Good luck with your further research!
Kind regards
