Flock Nocturnal Activity: Is There a Rotative Guard?

Round 1

Reviewer 1 Report

This study describes the nocturnal resting behaviour of 10 sheep, based on accelerometry. The study uses a dataset that was previously published by the authors. The authors present a summary of the number of times the sheep got up, their lying posture, the number of rest periods per night, and the average durations of rest and activity. The authors also present a plot of sheep activity over time, which appears to suggest that at least one sheep was active at any given time.

My impression of the current manuscript is that it presents some preliminary exploration of the data, but without clear objectives. The question in the title (‘is there a rotative guard’) is not mentioned in the study background or methods, and is not quantitatively analysed within the paper. The manuscript also requires more thorough proofreading and editing. For example, references are incorrectly numbered, there are still track changes within the manuscript, figures are missing axis labels, and the first sentence of Conclusions is repeated twice.

I provide some specific (but not comprehensive) suggestions for revisions below. I believe that the manuscript will require some reconsideration, restructuring, and more in-depth analyses before it is suitable for publication.

GENERAL COMMENTS

1. The focus of the study (sleep, active periods and lying habits) is broad and the objectives are vague. What do we specifically need to know about the sleep, active periods and lying habits of sheep, and why is this relevant or important? If the purpose is just to confirm whether accelerometry provides information about rest cycles that is consistent with the existing literature (based on more invasive methods), then this is fine. However, this needs to be more coherent throughout the manuscript.

2. Further analyses of the data could provide much greater insight. For example, to address the question of whether there is a rotating guard, you could quantify how many sheep are active at any given time of night. Is there actually at least one sheep active at any given time? The focus of the analyses will also, of course, depend on the focus of the manuscript.

3. The logical structure and flow of the manuscript requires some improvement. For example, there are currently comparisons with the existing literature, as well as descriptions of methods, in the Results section. These would be better placed in the Discussion and Methods sections, respectively.

4. It is worth acknowledging that sheep might sometimes (or even often) be inactive but alert.

INTRODUCTION / RELATED WORKS

5. It would be better to combine these two sections (Introduction and Related Works) into a single synthesis of the existing literature, to provide relevant context for the study.

MATERIALS AND METHODS

6. There is no need to provide so much detail about how the data were cleaned and processed prior to analysis. A single statement (e.g. ‘sheep with less than x hours of data were excluded from analyses’) would be sufficient.

7. Line 178: Please describe the statistical analyses that were performed (even if code is also provided).

FIGURES

8. Please add descriptive labels/titles to figure axes (e.g. time, overall tri-axial activity).

9. Figure 2 seems to duplicate information that is also presented in Figure 3. If so, it should be removed.

REFERENCES

10. Please check that all references are correctly numbered.

Author Response

Comments and Suggestions for Authors

This study describes the nocturnal resting behaviour of 10 sheep, based on accelerometry. The study uses a dataset that was previously published by the authors. The authors present a summary of the number of times the sheep got up, their lying posture, the number of rest periods per night, and the average durations of rest and activity. The authors also present a plot of sheep activity over time, which appears to suggest that at least one sheep was active at any given time.

My impression of the current manuscript is that it presents some preliminary exploration of the data, but without clear objectives. The question in the title (‘is there a rotative guard’) is not mentioned in the study background or methods, and is not quantitatively analysed within the paper. The manuscript also requires more thorough proofreading and editing. For example, references are incorrectly numbered, there are still track changes within the manuscript, figures are missing axis labels, and the first sentence of Conclusions is repeated twice.

We would like to start by thanking reviewer 1 for the thorough review and extensive list of suggestions he provided. And we also want to acknowledge the huge structuring and editing flaws the paper had.

I provide some specific (but not comprehensive) suggestions for revisions below. I believe that the manuscript will require some reconsideration, restructuring, and more in-depth analyses before it is suitable for publication.

The paper has been deeply edited, and we sincerely hope that we have successfully address the faults.

GENERAL COMMENTS

1. The focus of the study (sleep, active periods and lying habits) is broad and the objectives are vague. What do we specifically need to know about the sleep, active periods and lying habits of sheep, and why is this relevant or important? If the purpose is just to confirm whether accelerometry provides information about rest cycles that is consistent with the existing literature (based on more invasive methods), then this is fine. However, this needs to be more coherent throughout the manuscript.

We truly agree with the comment. The abstract, the introduction and the conclusions were edited to highlight the exploratory character of the paper, validating the technique and explaining the option of validating the various results documented in the literature.

2. Further analyses of the data could provide much greater insight. For example, to address the question of whether there is a rotating guard, you could quantify how many sheep are active at any given time of night. Is there actually at least one sheep active at any given time? The focus of the analyses will also, of course, depend on the focus of the manuscript.

We completely agree with the reviewer's argument, and we even recognize that it is not easy to prove it. In addition to not being able to verify that there is always a guard element, because as reviewer notices below, they may be guarding, but inactive. A second reason is since the experiment was carried out with a limited number of sheep with collars, and in the same period more animals spent the night in the same shelter. In this context, it is possible that guarding was carried out at certain times of the night by other unmonitored animals. In any case, rotating guard is a plausible explanation, and the paper does not claim the phenomenon occurs, it simply hypothesizes that it does.

Anyhow we processed figure 3 data in order to highlight the active and inactive periods, and created a new plot (figure 4).

The aim of the study was to validate inertial sensors as an inexpensive and non-invasive technique for assessing activity. While reading the state of the art I looked for more information about the alert states referred to in the literature and that exist in the rest cycle. I ended up finding a breeder's blog, which stated in a post that the sheep never all sleep at the same time and that they keep the flock in rotation. Obviously, we compared the activity of the various elements of the herd as soon as we could, because it is plausible hypothesis and it is quite interesting. Miserably I didn't save the blog address, and I still haven't found it again.

3. The logical structure and flow of the manuscript requires some improvement. For example, there are currently comparisons with the existing literature, as well as descriptions of methods, in the Results section. These would be better placed in the Discussion and Methods sections, respectively.

Paper was deeply restructured: introduction and related work sections were merged, and the comparisons were transferred to discussion sections, the descriptions were moved to methods section.

4. It is worth acknowledging that sheep might sometimes (or even often) be inactive but alert.

Yes, and we cannot detect it with the accelerometry. Anyhow since the plan is to infer energy consumption and to detect rest anomalies, we can continue to tune our monitoring algorithms. We added an acknowledging sentence.

INTRODUCTION / RELATED WORKS

5. It would be better to combine these two sections (Introduction and Related Works) into a single synthesis of the existing literature, to provide relevant context for the study.

Yes, we did the merging.

MATERIALS AND METHODS

6. There is no need to provide so much detail about how the data were cleaned and processed prior to analysis. A single statement (e.g. ‘sheep with less than x hours of data were excluded from analyses’) would be sufficient.

It was simplified.

7. Line 178: Please describe the statistical analyses that were performed (even if code is also provided).

We added the following paragraphs that summarizes the statistical analysis.

The statistical analysis is straightforward. Based on the activity value computed by equation 1 we define a filtering system based on two thresholds. For each sheep, and for each night we select the activity value and compute the 40% and 80% percentile. Any activity value bellow 40% is considered asleep, any activity value above 80% is considered awake. Activity values in between are soft activity or sleeping and where not considered. The thresholds were defined by manual inspection of a sample from the dataset. It is important to mention that two smoothing filters are applied based on Exponential Moving Average to account for the rash movements and sensitivity of the sensors.

We followed a similar procedure to analysis from which side a sheep gets-up. Instead of using the previous mention thresholds, we consider the accelerometer roll with the range of [-10, 10] degrees. Any value that exceeds the range is classified as left or right, respectively. If the roll value is within the range, it is classified as center.

Finally, we measure the time each sheep spends in each activity state and on each direction to compute the statistics presented on the document.

FIGURES

8. Please add descriptive labels/titles to figure axes (e.g. time, overall tri-axial activity).

We added labels to the plots to facilitate the understanding of the materials.

Thanks for the comment.

9. Figure 2 seems to duplicate information that is also presented in Figure 3. If so, it should be removed.

Figure 2 visually adds information that is not possible to show in Figure 3. In order not to repeat the information, we chose to analyze the joint behavior of the herd on a different night (05/12/2021).

REFERENCES

10. Please check that all references are correctly numbered.

They were in fact not correctly ordered, but it was fixed in the text.

Author Response File: Author Response.docx

Reviewer 2 Report

In the manuscript entitled ‘Flock nocturnal activity: is there a rotative guard?’, the authors analyzed a public dataset created with data gathered by wearable sensors applied on sheep during the nocturnal period. The authors focused on the rest-activity of sheep, and analyzed their sleep and active periods and also their lying habits. Overall, monitoring animal activity during night is interesting and important to understand their survival conditions and providing supporting information for livestock farming. This study is mostly descriptive and provides a very general idea on the cyclical character of nocturnal activity and the typical posture when lying down of sheep, and therefore, the merit of this study is limited. I have some concerns and suggestions on both analytical and scientific aspects.

(1) As mentioned in the conclusion, guard rotation is just an assumption and lacking of powerful evidence from the results of this manuscript, a more appropriate title should be chosen rather than ‘Flock nocturnal activity: is there a rotative guard?’. Although the present title is interesting and attractive, however, not in accordance with the content of the manuscript.

(2) Line 160: There is a missing of units on the y-coordinate in Figure 1. And the sentence ‘Most collars recorded the animal activity for more than 380 hours (see Figure 1)’, could not in accordance with the Figure 1, especially without the description of the data conversion method referred to 380 hours and the values of the bar in Figure 1.

(3) Latin names should be listed after Serra da Estrela sheep.

Repetitive sentences should be simplified, for example. Line 150 and Line 156 ‘between November 18, 2021, and February 16, 2022’. Please check throughout this manuscript.

(4) It’s confused that the efficient data used for final analysis was obtained from the 10 collars or the 10 sheep? The number records (x-component) showed in Figure 1 represented the collars (also Line 166), however, they represented the sheep ID in Table 1 and Table 2. Which is the real record? Please check and correct.

(5) More detail information should be added in the ‘3. Nocturnal activity monitoring’, which is a similar section like ‘Materials and Methods’. For the sheep, where to wear the collars? What the model parameters and manufacturer of the collar used in this study? Is there a picture of the collar and the picture of the experimental sheep? Are these collars worn all the time or only at night? If only night, what time begin and end? Is there a sexual distinction of the sheep used in this experiment? if yes, how many males or females?

(6) Line 178: ‘the data were then statistically analyzed’, please describe the statistical methods more detailly and clearly.

(7) Several factors may affect the results of the experiment, like the density of sheep and environmental factors (temperature, humidity, rainfall, noise volume level, etc.), these should be taken into account for the data analyses. If not, at least to specify the limitation of the sampling scheme.

Author Response

Comments and Suggestions for Authors

In the manuscript entitled ‘Flock nocturnal activity: is there a rotative guard?’, the authors analyzed a public dataset created with data gathered by wearable sensors applied on sheep during the nocturnal period. The authors focused on the rest-activity of sheep, and analyzed their sleep and active periods and also their lying habits. Overall, monitoring animal activity during night is interesting and important to understand their survival conditions and providing supporting information for livestock farming. This study is mostly descriptive and provides a very general idea on the cyclical character of nocturnal activity and the typical posture when lying down of sheep, and therefore, the merit of this study is limited. I have some concerns and suggestions on both analytical and scientific aspects.

 We would like to begin by thanking reviewer 2 for his careful review and extensive list of suggestions he provided.

(1) As mentioned in the conclusion, guard rotation is just an assumption and lacking of powerful evidence from the results of this manuscript, a more appropriate title should be chosen rather than ‘Flock nocturnal activity: is there a rotative guard?’. Although the present title is interesting and attractive, however, not in accordance with the content of the manuscript.

 We agree with reviewer 2 that we have not proved the existence of the rotating guard mechanism, because we have a very small sample of animals, because it is possible that the animals are alert even if inactive, and because we cannot guarantee that there are other animals staying overnight. in the same space without being monitored. But the paper is limited to proposing this hypothesis, which is possible, and which we plan to study in more detail in the future, solving the problems identified above.

The rotating guard hypothesis is not described in the introduction due to the conditions under which the work was carried out. In fact, the initial purpose of the work was to validate the inertial sensors as a way of inferring the energy consumed during the night period, and to be able to feed a machine learning engine that optimizes animal feeding. During the analysis of the state of the art, we found a blog of a creator who mentioned it, and we compared the activity temporally. In short, it is an unexpected result of the work we have done.

(2) Line 160: There is a missing of units on the y-coordinate in Figure 1. And the sentence ‘Most collars recorded the animal activity for more than 380 hours (see Figure 1)’, could not in accordance with the Figure 1, especially without the description of the data conversion method referred to 380 hours and the values of the bar in Figure 1.

That’s true, 380 hours is not the correct value and we edited in the text.

(3) Latin names should be listed after Serra da Estrela sheep.

It was added in the text.

Repetitive sentences should be simplified, for example. Line 150 and Line 156 ‘between November 18, 2021, and February 16, 2022’. Please check throughout this manuscript.

Something strange happened, we believe because of tracking changes created between different editors. The repeated sentences were removed.

(4) It’s confused that the efficient data used for final analysis was obtained from the 10 collars or the 10 sheep?

During the dataset creation, we always used the same collar on each sheep, even though they were removed to charge batteries. That's why we use sheep and collar identification indiscriminately. In order to avoid any doubt, we edited the text and tables to always identify the animal through collar ID.

The number records (x-component) showed in Figure 1 represented the collars (also Line 166), however, they represented the sheep ID in Table 1 and Table 2. Which is the real record? Please check and correct.

Sheep used in the monitoring process were identified in the dataset by the collar ID. They wore the same collar throughout the entire experience. As such, in the present paper, the identification of the animal continues to be through the collar number. In the first data verification performed, we realized that some of the necklaces reported little data, which is why we did not consider them in the analysis. Therefore, the activity data of collars 1, 5, 8, 10 and 15 are not represented in the graphs. The collar numbers used in the y-axis of figure 3 correspond to the values of the xx-axis of figure 1. In order not to confuse the reader, the text and tables were edited to emphasize that the identification depends on collar ID.

(5) More detail information should be added in the ‘3. Nocturnal activity monitoring’, which is a similar section like ‘Materials and Methods’. For the sheep, where to wear the collars?

We changed the section names, and we added a URL of the product webpage as a footnote.

What the model parameters and manufacturer of the collar used in this study?

The collar model and vendor were referred in the paper, and it was included as well as a reference to a paper that added more detailed information about the dataset, the collars and the monitoring process. Paper was not referred in the initial version of current paper, since it was not available on the submission time.

 Is there a picture of the collar and the picture of the experimental sheep?

The footnote URL includes pictures about the collars and animals wearing them. We hope it will be enough for readers who want to understand the dimensions and shape of the collar, to be able to do so.

Are these collars worn all the time or only at night?

Collars were worn 24 hours a day, but the data was just gathered at sheepfold, since there was no radio coverage during the pasture period. The new reference [31] adds information on it.

If only night, what time begin and end?

The analysis just used records gathered between 18:00 and 9:00, and the text was edited to highlight the time interval. The new reference [31] adds information on it.

 Is there a sexual distinction of the sheep used in this experiment? if yes, how many males or females?

In the experience they were monitored just females. It was added a remark about it in the text.

(6) Line 178: ‘the data were then statistically analyzed’, please describe the statistical methods more detailly and clearly.

 We added the following paragraphs that summarizes the statistical analysis.

The statistical analysis is straightforward. Based on the activity value computed by equation 1 we define a filtering system based on two thresholds. For each sheep, and for each night we select the activity value and compute the 40% and 80% percentile. Any activity value bellow 40% is considered asleep, any activity value above 80% is considered awake. Activity values in between are soft activity or sleeping and where not considered. The thresholds were defined by manual inspection of a sample from the dataset. It is important to mention that two smoothing filters are applied based on Exponential Moving Average to account for the rash movements and sensitivity of the sensors.

We followed a similar procedure to analysis from which side a sheep gets-up. Instead of using the previous mention thresholds, we consider the accelerometer roll with the range of [-10, 10] degrees. Any value that exceeds the range is classified as left or right, respectively. If the roll value is within the range, it is classified as center.

Finally, we measure the time each sheep spends in each activity state and on each direction to compute the statistics presented on the document.

(7) Several factors may affect the results of the experiment, like the density of sheep and environmental factors (temperature, humidity, rainfall, noise volume level, etc.), these should be taken into account for the data analyses. If not, at least to specify the limitation of the sampling scheme.

The limitation was identified, and it was defined as future work.

Author Response File: Author Response.docx

Reviewer 3 Report

The investigations are very promising, done by modern technology. The interpretation of the results is not suitable for scientific publicatuion in the present form. It should be improved. Recommended to form groups of animals to evaluate their activity, or to calculate average resting, moving etc. time to compare the data with literature. Very few data are presented in the paper from the total. We really don't know what is happening between 18 November 2021 and 16 February 2022 with the flock. Calculation of sum or/and average data is recommended.
Reviewing the text by native English speaker is recommended, as well. Explanations of the data and the discussion part are confusing. Sometimes, less is more...

Detailed comment and suggestions can be found in the text, but the reviewer did not make many changes, becouse there are a lot of grammatical and professional mistakes.

Comments for author File: Comments.pdf

Author Response

The investigations are very promising, done by modern technology. The interpretation of the results is not suitable for scientific publicatuion in the present form. It should be improved.

We would like to thank reviewer 3 for the reviewing effort and suggestions.

Recommended to form groups of animals to evaluate their activity, or to calculate average resting, moving etc. time to compare the data with literature. Very few data are presented in the paper from the total.

We appreciate the comment and agree that it will make perfect sense when we have more animals represented in the study, as clustering would further decrease the variance of the analysis. It is important to note that the analysis is global (the same methods and filters are applied to all sheep), however, the results are accounted per sheep. We are already collecting a new dataset with a larger volume of sheep. The new dataset will be analyzed with your suggestion in mind. This line of work has been added to the conclusions section.

We really don't know what is happening between 18 November 2021 and 16 February 2022 with the flock. Calculation of sum or/and average data is recommended.

While the images are merely illustrative of a single night and present a graphical way to understand the analysis. The tables presented contain the average of the crucial characteristics we wanted to study and present in this work. We would like to point you to table 2, which was enriched with a suggestion for another reviewer.

Reviewing the text by native English speaker is recommended, as well. Explanations of the data and the discussion part are confusing. Sometimes, less is more...

We reviewed and edited the document to polish the language used.

Detailed comment and suggestions can be found in the text, but the reviewer did not make many changes, becouse there are a lot of grammatical and professional mistakes.

Thanks for the suggestion, we tried to clean as many issues as possible in this new version of the document.

-- FROM reviewer 3 PDF annotations:

Line 100.2014: “It is not a common way to describe the content of the paper. Not needed, recommended to delete.”

We would like to thank the reviewer for the comment, we edited the text accordingly.

Line 109: “There is not enough information about the keeping system and animal rearing (indoor or outdoor; grazing or hand-feeding etc.).

The paper analyses a public dataset, and we referenced in [27] the paper that statically describes the data and the way it was gathered. We just avoided repeating the information here in order not to bother the reader.

Also, data about the macro- and micro-climatic circumstances are missing. Climate (mostly the temperature and the humidity - THI) has a great effect on the activity of the animals.”

We thank the suggestions, but unfortunately, that kind of data was not gathered by the monitoring system. Anyhow the animals spend the night at a sheepfold. We certainly will request the weather data source for the future analysis of the novel dataset that is currently being gathered.

Line 110: “Ovis aries, not ovis aries.”

Thanks for the comment, we fixed it based on your suggestion.

Line 158: “Feeding times, milking times and other acivities (e.g. cleaning and other stockpersons' movements) are not mentioned in the text. Are there any of them after 6 p.m.? Perhaps these effects can couse some of the peaks on Figure 2.”

The animals are collected, milked, and taken to a closed compartment at 18:00. After the space has been prepared and the food has been placed there. The animals are within that closed compartment until the handler returns the next morning, at which time he opens them and takes them to be milked, and later puts them on pasture. During the night there is no human intervention, and since the space is in an uninhabited place, there is no disturbance of any kind. The paper was edited, and a note was added.

Line 164: “No need to explain.”

The extra information was added after a request from another reviewer.

Line 210: “No need to explain.”

The extra information was added after a request from another reviewer.

Linw 226: “No need to explain.”

The extra information was added after a request from another reviewer.

Lin2 238: “"Members"? Individuals, animals or something like that.”

Thanks for the comment, w

Round 2

Reviewer 1 Report

Thank you to the authors for their revisions to the manuscript. Although I find the manuscript somewhat improved, there are still fundamental and critical issues with this study. In particular, I do not believe that the conclusions are supported by the results of the study, the methods do not adequately address the supposed aims, and the analysis of the data is still only rudimentary. I also have concerns about the validity of the authors’ thresholds for wake and rest. I provide detailed comments below. Once again, these comments are not comprehensive, as I have only focussed on the most major issues with the manuscript. There are still many more minor issues with the manuscript that would need to be addressed.

*MAJOR COMMENTS*

1. The abstract and introduction still do not accurately represent the content of the manuscript. You state that the present study intends to “validate non-invasive techniques as a way to infer energy expenditure at night, and to detect abnormal nocturnal activity”. However, energy expenditure is hardly mentioned in the study (although the activity measure used could be used as a proxy for energy expenditure) and there is no attempt to characterise overall energy expenditure across the night. Nor does the study provide any definition or analysis of normal versus abnormal activity. Please revise these sections to provide a clear statement of the present study’s scope and purpose. In your response to one of my comments, you stated that “The aim of the study was to validate inertial sensors as an inexpensive and non-invasive technique for assessing activity”, which seems like a more reasonable and accurate description.

2. I appreciate that there are some limitations to your dataset. Nevertheless, if the goal here is to validate non-invasive techniques for monitoring activity, it seems reasonable to expect further analyses that attempt to determine whether the method is valid and useful. So far, this paper presents a general description of sheep activity and behaviour. Would you be able to diagnose, from these data alone, any potential issues with a particular sheep? Was the variation that you observed real or an artefact of the collars? How does the periodicity of rest cycles that you observed compare with sleep cycles recorded using more invasive methods? I would like to see some further consideration of these issues in the exploration of the data, as well as in the discussion.

3. How did you perform the “manual inspection of a sample from the dataset” to decide on thresholds for sleeping and wakefulness? Did you observe the behaviour of the animals, or just the data? Did you confirm these thresholds for multiple individuals? When looking at Figure 2, I do not find these thresholds convincing. There is a lot of variation in activity within the ‘sleep’ bracket, and it seems that for a large proportion of the animals’ time, it is undefined whether the animal is active or resting.

4. The data presented in Figure 4 do not match the general conclusions of your study. Rather than seeing cycles of rest and activity, here we see that some individuals rested for almost the entire night, while another was (as you put it) seemingly by some collar error active the entire night. Instead of validating the use of these collars for monitoring, these data seem to present issues with the definitions of rest and activity and/or the time intervals used when consolidating the data. These need to be resolved before anyone can say that this is a reliable method for monitoring animal activity at night.

5. The fundamental question of whether your results are consistent with the results of previous studies is not actually clearly addressed in your Discussion or Conclusions. You state that the values are similar, but do not provide these values for comparison, and also note that there is a high degree of variability in your study. Consequently, the results of this study alone do not give me any confidence in the reliability of this method for measuring or diagnosing issues with nocturnal activity.

6. Lines 285-288: Where are the data to support these conclusions? I have not seen anything to clearly support this conclusion anywhere within this manuscript.

*MINOR COMMENTS*

GENERAL

7. Please note that I have not provided comments for English langugage editing. While the contents of the manuscript can generally be understood, there are still many issues with spelling, grammar and syntax throughout the text. There are also some passages that are technically correct but described in an unusual way for a scientific journal (e.g. Lines 105-108). I suggest asking somebody else to assist with editing before publication.

TITLE

8. The title needs to be revised to more accurately represent the content of the manuscript. I suggest: “Monitoring nocturnal resting behaviour in sheep using accelerometry”.

INTRODUCTION

9. The introduction is improved from the previous version, but still contains a series of detailed summaries of previous studies, each in a separate paragraph. This is difficult to read and the purpose or relevance is not always clear. Please consolidate this information such that each paragraph makes a key point or topic that is relevant to the present study.

MATERIALS AND METHODS

10. I presume that Dx, Dy, Dz refer to the dynamic acceleration of each axis (with static acceleration removed) rather than the raw acceleration? Please clarify. It also seems that the activity measure you have used is referred to as vectorial dynamic body acceleration (VeDBA) in other studies (e.g. Wilson et al. 2019: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7030956/) so it would be helpful to use this term throughout.

11. Line 157: “Compute the statistics” is very vague. It is actually unclear in your results whether you are referring to standard error or standard deviation. Please provide a clearer statement, e.g. “We calculated the average (± standard error) time that sheep spent in each state and that they spent lying on each side”

RESULTS

12. The Results section should only present the results of the study, not how these were obtained. There are still many descriptions of methods in the Results section, which should be moved to the Materials and Methods section (e.g. Lines 170-172). Please check the manuscript again carefully.

13. In Table 2, the columns are labelled “Rest time per night (min)” and “Active time per night (min)”, but the text states that these are “the number of rest cycles and their duration”. Please clarify whether these values are per rest cycle or per night. As mentioned by the other reviewer, it would be clearer to refer to sheep ID rather than collar ID, as well.

14. If the intention is to compare with the existing literature, then the measures reported should be comparable with the existing literature. For example, in the discussion, you cite total inactivity at night for other studies, but there is no measure of total inactivity at night for the present study.

DISCUSSION

15. In the first paragraph of the discussion, you report values from other studies and state that the period of inactivity found in the present study is similar. Please clarify this section so that all data is in the same units and is comparable. Furthermore, throughout your study, you describe the durations of inactivity per rest cycle but not across the entire night. This would be useful to include, for comparison.

Author Response

Thank you to the authors for their revisions to the manuscript. Although I find the manuscript somewhat improved, there are still fundamental and critical issues with this study. In particular, I do not believe that the conclusions are supported by the results of the study, the methods do not adequately address the supposed aims, and the analysis of the data is still only rudimentary. I also have concerns about the validity of the authors’ thresholds for wake and rest. I provide detailed comments below. Once again, these comments are not comprehensive, as I have only focused on the most major issues with the manuscript. There are still many more minor issues with the manuscript that would need to be addressed.

We would like to start by thanking reviewer 1 for the thorough review and for the extensive list of comments he provided.

*MAJOR COMMENTS*

1. The abstract and introduction still do not accurately represent the content of the manuscript. You state that the present study intends to “validate non-invasive techniques as a way to infer energy expenditure at night, and to detect abnormal nocturnal activity”. However, energy expenditure is hardly mentioned in the study (although the activity measure used could be used as a proxy for energy expenditure) and there is no attempt to characterise overall energy expenditure across the night.

The test was edited both in the discussion and conclusion sections and now it clearly states our limitations and our analysis of the data quality, we also added how the data collection could be enhanced in a feeding optimization process. It would be possible to estimate the various components of sleep according to Toutain and to perform an approximate calculation of the energy consumed in the period, we did not do it because the dataset does not include the weight of the animals at the time of the experiment and calculate it with the present weight would be neither reasonable nor useful. Anyhow nocturnal activity data might be helpful to feed a machine learning process enabled with feeding, weighting, and milk production information, as we wrote in the conclusions.

Nor does the study provide any definition or analysis of normal versus abnormal activity. Please revise these sections to provide a clear statement of the present study’s scope and purpose. In your response to one of my comments, you stated that “The aim of the study was to validate inertial sensors as an inexpensive and non-invasive technique for assessing activity”, which seems like a more reasonable and accurate description.

Our aim was to validate the inertial sensors as a way to assess activity since animal activity consumes energy, and impacts optimal feeding. Our interest does have to do with the amount of Joule's animal spent during the night, or during the entire day, but to understand the optimal amount of aliment it needs to keep its optimal production. That description was added in sections 4 and 5.

2. I appreciate that there are some limitations to your dataset. Nevertheless, if the goal here is to validate non-invasive techniques for monitoring activity, it seems reasonable to expect further analyses that attempt to determine whether the method is valid and useful. So far, this paper presents a general description of sheep activity and behaviour. Would you be able to diagnose, from these data alone, any potential issues with a particular sheep?

We would like to thank the reviewer for the comment, we did not mention previously, but alongside the dataset, we also have access to two weeks of footage from the sheeps. The footage was used to manually verify the validity of the values gathered from the collars as well as the correctness of the statistical analysis.

Was the variation that you observed real or an artefact of the collars? How does the periodicity of rest cycles that you observed compare with sleep cycles recorded using more invasive methods? I would like to see some further consideration of these issues in the exploration of the data, as well as in the discussion.

We are not sure we understand the question well, but the animals were monitored by a collar during the period we announced, and they were video recorded during part of that period. The dataset we use is public, it includes the data generated by the collar and the video recordings, and it is possible to visually verify the behavior of the animals, and even compare the videos with the accelerometry data created by the collars. In the paper we describe how we analyzed the data, we share the analysis code, and we show the results of the number of cycles of activity per night, the average times of inactivity and activity in each cycle, and in the present version, we also add the total time of inactivity and activity per night. In the paper, we also compared the values obtained with the values reported in the literature and show that the values reported in the literature are within the error range of our study. We sincerely hope to have answered the question, but we confess that we are not sure we have understood the question.

3. How did you perform the “manual inspection of a sample from the dataset” to decide on thresholds for sleeping and wakefulness? Did you observe the behaviour of the animals, or just the data? Did you confirm these thresholds for multiple individuals? When looking at Figure 2, I do not find these thresholds convincing. There is a lot of variation in activity within the ‘sleep’ bracket, and it seems that for a large proportion of the animals’ time, it is undefined whether the animal is active or resting.

As previously stated, we used two weeks of footage to validate the threshold from the statistical analysis. It is also important to point out, that the threshold is applied equally to all sheep, and we did not compute thresholds for each individual sheep. The focus of the work is to show that external monitorization of sheep could yield the same results as other (more intrusive) methods. With that in mind, and that we have a rather limited dataset, we applied a straightforward statistical analysis to show the potential of the method. Going further with this work we intend to improve the data acquisition method, as well as the analysis done on the data. This information was added in section 2.

4. The data presented in Figure 4 do not match the general conclusions of your study. Rather than seeing cycles of rest and activity, here we see that some individuals rested for almost the entire night, while another was (as you put it) seemingly by some collar error active the entire night.

Instead of validating the use of these collars for monitoring, these data seem to present issues with the definitions of rest and activity and/or the time intervals used when consolidating the data. These need to be resolved before anyone can say that this is a reliable method for monitoring animal activity at night.

The images are simply illustrative, while the tables provide the aggregated data from all the valid nights.

The images represent the activity of the animals in one night, and the night was chosen because it is illustrative of animal behavior and because it is the most complete data sequence with all collars at the same time.

Concerning Figure 4, it shows that there is an initial disturbance of collars 4, 9, 17, and 20, which conditioned the analysis on that day. This disturbance may have been simply a shock, which may even have occurred between two animals. This intense change in activity conditioned the threshold (which is defined by the percentile) thus limiting the sensitivity of what should correspond to the active cycle and the inactive cycle. As can be seen from the small activity peaks and their temporal separation detected by collars 4, 9, 17, and 20, the activity pattern observed in the other animals can be verified in those collars, but with a smaller amplitude. We decided to not cherry-pick the data to create the images. For instance, by removing the formerly referred activity peaks, we would have perfect behavior for the referred collars in the images.

The averages of total active and inactive time in each of the 53 monitored nights, that was added to the current version of the paper, produce values in line with Toutain and Zobber publications. And this is what impacts energy spent, and what must be considered in the establishment of an activity indicator to optimize the feeding process.

5. The fundamental question of whether your results are consistent with the results of previous studies is not actually clearly addressed in your Discussion or Conclusions. You state that the values are similar, but do not provide these values for comparison, and also note that there is a high degree of variability in your study. Consequently, the results of this study alone do not give me any confidence in the reliability of this method for measuring or diagnosing issues with nocturnal activity.

The figures are merely illustrative of the results (they represent a maximum of one night of activity) while the tables capture the aggregate of the entire time of the experiment. The tables have been extended to add the average activity and inactivity totals for each night. The results shown in the tables include the idle and active time values that are within the margins given by Zobel and Toutain, as written in 206-207.

6. Lines 285-288: Where are the data to support these conclusions? I have not seen anything to clearly support this conclusion anywhere within this manuscript.

The estimate of activity based on accelerations was compared on line 206-207 with Toutain and Zobel and their results lie within our confidence intervals. Table 2 was added with 2 new columns to show emphasize the results are in line with the literature.

*MINOR COMMENTS*

GENERAL

7. Please note that I have not provided comments for English langugage editing. While the contents of the manuscript can generally be understood, there are still many issues with spelling, grammar and syntax throughout the text. There are also some passages that are technically correct but described in an unusual way for a scientific journal (e.g. Lines 105-108). I suggest asking somebody else to assist with editing before publication.

We carried out a general edition of the text.

TITLE

8. The title needs to be revised to more accurately represent the content of the manuscript. I suggest: “Monitoring nocturnal resting behaviour in sheep using accelerometry”.

In our humble opinion, the title of the paper should be indicative of its content, it should be realistic regarding the results it presents, and it should be appealing. We consider that the conditions must be verified in this order. Finally, we consider that the title we have chosen for this paper meets all the conditions. We respect your opinion; but please allow us to disagree.

INTRODUCTION

9. The introduction is improved from the previous version, but still contains a series of detailed summaries of previous studies, each in a separate paragraph. This is difficult to read and the purpose or relevance is not always clear. Please consolidate this information such that each paragraph makes a key point or topic that is relevant to the present study.

We edited the text, and we truly believe that we improved it.

MATERIALS AND METHODS

10. I presume that Dx, Dy, Dz refer to the dynamic acceleration of each axis (with static acceleration removed) rather than the raw acceleration? Please clarify. It also seems that the activity measure you have used is referred to as vectorial dynamic body acceleration (VeDBA) in other studies (e.g. Wilson et al. 2019: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7030956/) so it would be helpful to use this term throughout.

We thank the reviewer for the suggestion and the reference. The text was edited to clarify the metrics used to analyze animal activity. The reviewer is right, we used VeDBA, which is not suitable for energy analysis, but given the exploratory nature of our work, we decided to keep it, since it is more suitable for animal motion analysis. As we wrote in the paper, we plan to continue to work on energy balance and we will compute the dataset that is being created, using the ODBA metric.

11. Line 157: “Compute the statistics” is very vague. It is actually unclear in your results whether you are referring to standard error or standard deviation. Please provide a clearer statement, e.g. “We calculated the average (± standard error) time that sheep spent in each state and that they spent lying on each side”

We edited the sentence accordingly.

RESULTS

12. The Results section should only present the results of the study, not how these were obtained. There are still many descriptions of methods in the Results section, which should be moved to the Materials and Methods section (e.g. Lines 170-172). Please check the manuscript again carefully.

The text just refers to the interpretation of the image, not to the analysis or the data-gathering process. It was just added at the request of another reviewer. It also serves to ensure the reproducibility of all results published in this paper.

13. In Table 2, the columns are labelled “Rest time per night (min)” and “Active time per night (min)”, but the text states that these are “the number of rest cycles and their duration”. Please clarify whether these values are per rest cycle or per night. As mentioned by the other reviewer, it would be clearer to refer to sheep ID rather than collar ID, as well.

Cycles were counted per night; the average inactive and active periods were measured per cycle. We edited the text to highlight it and we created new columns in table 2.

14. If the intention is to compare with the existing literature, then the measures reported should be comparable with the existing literature. For example, in the discussion, you cite total inactivity at night for other studies, but there is no measure of total inactivity at night for the present study.

We added the information in table 2.

Author Response File: Author Response.pdf

Reviewer 2 Report

Overall the manuscript has been improved in that the manuscript now somewhat more accurately represents the truth about the nocturnal activity of sheep. The authors have answered and revised all my concerns. There are still several details of the figure and table need to be improved.

Figure 2 is not cited in the main text. The title of Table 2 is too crude. And it would be better to polish up the language to clearly express the research findings.

Author Response

Comments and Suggestions for Authors

Overall the manuscript has been improved in that the manuscript now somewhat more accurately represents the truth about the nocturnal activity of sheep. The authors have answered and revised all my concerns. There are still several details of the figure and table need to be improved.

We would like to thank reviewer 2 for the reviewing effort and suggestions.

Figure 2 is not cited in the main text. The title of Table 2 is too crude.

We edited the text to make it more expressive.

And it would be better to polish up the language to clearly express the research findings.

We revised the all text.

Author Response File: Author Response.pdf