A Novel Scouring Method to Monitor Nocturnal Mammals Using Uncrewed Aerial Vehicles and Thermal Cameras—A Comparison to Line Transect Spotlight Counts

Round 1

Reviewer 1 Report

The authors present an overall well written and scientifically sound exploration of novel drone survey methodology for cryptic wildlife that is likely to be of high interest to ecologist and wildlife management decision makers. I have some concerns about the research design to allow comparison of the novel drone survey method and traditional spotlighting methods, and I feel discussion of the results of the study is lacking in detail in some places. I believe these points can be addressed by the authors making minor revisions to the methods, results and discussion sections. Please see attached reviewer comments to author document for details.

Comments for author File: Comments.pdf

English language is overall of high quality, though the manuscript would benefit from proofreading for infrequent grammatical errors/ 

Author Response

Dear reviewer 1

Thank you for your time, effort, and perspectives. We have addressed your comments and made the appropriate changes to the manuscript, please see below.

Best regards,

Peter, Dan, Cino, and Sussie

General comments:

Comment 1: The decision to conduct spotlight surveys first and drone surveys shortly afterwards when it is known spotlighting was like to disturb some of the species surveyed is concerning. The authors do well to address the impact this had on counts of deer and carnivores, however further justification for this survey design is needed.

Reply: This has been clarified in line 293-299: Since this study was based on the premise of the Hunters Association's planned surveys, it was a priority not to risk disturbing their surveys. Therefore, the drone surveys had to be conducted after the spotlight surveys. Ideally, the drone survey would have been conducted first, as the data suggests that spotlighting disturbs animal distribution. Even better would have been to conduct the surveys on separate nights. However, this approach would require many more surveys to achieve a normal distribution of data for both methods, providing a foundation for future studies.

Comment 2: The authors stated that the drone survey method showed potential to surpass spotlighting in terms of efficiency, however the method section states the drone surveying took approximately the same amount of time as spotlighting, and there is no information given about the time required to complete analysis of the footage. More information is needed to back up this claim, or if the methods are similar in time requirements than discussion should focus on the increased accuracy of drone surveying in comparison to spotlighting and other traditional methods for the same time investment.

Reply: This has been clarified in line 249-256: The drone method measured up to the traditional spotlighting method and showed potential to surpass it in terms of time spent, accuracy and applicability, when considering the additional area covered outside the transect area, as well as the likelihood that near all animals in open areas were observed by the drone inside the transect area. The time spent subsequently plotting the observations and analyzing footage were not directly compared between the two methods, but the spotlighting data were less time consuming to process. However, with more experience with the drone method and implementation of machine learning techniques, the difference could become negligible.

Comment 3: Further discussion is needed around the linear relationships found in the paper. It is stated that they are statistically significant, but the direction and strength of these relationships and what that means in terms of when drone surveying can be relied upon to produce accurate population estimates is not fully explained.

Reply: This has been clarified in line 274-278: The positive slopes of the linear regressions (Figure 3) for hare and carnivore were below 1, which suggests that the drone method in general has more observations than the spotlight method for these two groups. However, the p-values are relatively high for these two groups. For deer the slope is close to 1, which suggest little to no difference.

Comment 4: It would be interesting to see how spotlighting estimates adjusted using distance sampling methods compare to drone counts as well as the raw spotlighting counts. As if drone and distance sampling adjusted spotlight counts are very similar, an argument could be made that use of spotlighting should be preferred due to the low amount of resources and expertise needed to conduct these kinds of surveys.

Reply: Agreed. However, one thing to consider is that distance sampling assumes that all animals are spotted at distance 0 from the transect line, and the probability of detection decreases with the distance from the observer. The drone method, to some degree, assumes that all animals are spotted in the entire transect. Here the probability of missing animals would increase with the distance from the drone, outside the transect area, so in this setup distance sampling would not be directly comparable. But it would be very interesting to examine in a new study.

Specific comment 1: Figure 2 - I suggest replacing this with a version that is just a map, it is not necessary for the reader to see the software user interface.

Reply: Figure 2 has been cropped to show only the map

The manuscript has been proofread for grammatical errors.

Reviewer 2 Report

The paper entitled "A novel scouring method to monitor nocturnal mammals using UAVs and thermal cameras - a comparison to line transect spotlight counts" investigates a comparison between transect spotlight count surveys conducted by hunter association and a scouring approach with the drone with a thermal camera. The study is well articulated in the various paragraphs: the introduction explains the state of the art of drone monitoring applied to wildlife. It introduces the methodological section of the research carried out in northern Denmark in the spring of 2023. The methodology, carefully described, was made by comparing the two monitoring methods along predetermined transects at a short distance from each other. The results, presented with in-depth statistical analyses, show significant differences between the two methods as regards the hare (in favor of the drone approach). In comparison, the same considerations cannot be made for deer and carnivores (the results of the two methods are similar).

I believe the different behavior of the animals (hares versus deer+carnivores) is a response to the anthropic disturbance caused by the spotlight counts: the hares do not seem disturbed by the consecutive detection via spotlight counts and drone; the opposite seems to happen for deer and carnivores, which are present in the areas outside the transect. To not underestimate the presence of deer and carnivores, it would be necessary to consider the animals recorded by the thermal camera in the areas outside the transect or to reformulate the methodological approach by considering the two surveys at two different time points (for example, separate days). This could be the only weakness of the paper. In this sense, even the studies cited as a comparison (e.g., Koala by Witt et al.) confirm a greater probability of detection in drone + thermal camera (or machine learning) applications than traditional census methods.

These methodological approaches can be expensive regarding hardware and software instrumentation, but they represent a future challenge in monitoring wildlife, especially for cryptic animals. Another strong point is represented by monitoring large and difficult-to-investigate areas with less effort in terms of time. I am not saying that the spotlight counts should be replaced, but it will undoubtedly have to be supported by innovative and more objective methodologies. In Italy, for example, the hunter population is decreasing year after year, and therefore, hunting management, both for game bags and censuses, is becoming a problem.

The article can be accepted in the following form; I await a comment from the authors on the correct methodology to avoid mutual causes of disturbance between the two methods.

Author Response

Dear reviewer 2

Thank you for your time, effort, and perspectives. We have addressed your comments and made the appropriate changes to the manuscript, please see below.

Best regards,

Peter, Dan, Cino, and Sussie

Comment 1: To not underestimate the presence of deer and carnivores, it would be necessary to consider the animals recorded by the thermal camera in the areas outside the transect or to reformulate the methodological approach by considering the two surveys at two different time points (for example, separate days). This could be the only weakness of the paper.

The article can be accepted in the following form; I await a comment from the authors on the correct methodology to avoid mutual causes of disturbance between the two methods.

Reply: This has been clarified in line 286-292: Since this study was based on the premise of the Hunters Association's planned surveys, it was a priority not to risk disturbing their surveys. Therefore, the drone surveys had to be conducted after the spotlight surveys. Ideally, the drone survey would have been conducted first, as the data suggests that spotlighting disturbs animal distribution. Even better would have been to conduct the surveys on separate nights. However, this approach would require many more surveys to achieve a normal distribution of data for both methods, providing a foundation for future studies.

The manuscript has been proofread for grammatical errors.