Sheep Nocturnal Activity Dataset
Round 1
Reviewer 1 Report
the manuscript presents a very robust and well collected set of data on nocturnal activity of sheep. the presentation, although not excellent, adequately contextualizes the data in the scientific state-of-the-art.
However, the paper could benefit from some remodelling, especially in section 2. It should start with a comprehensive methodology for site selection, data collection and analysis/processing, and then follow-up with the data summary. the 3-step presentation at the start of 2.2 informs very little of the steps, how they were performed, it mentions nothing on precautions and back-up plans for the case of data collection or transfer tools presenting problems, as well as not explaining how the "malformed" records are detected, why they are considered so, and needs to be clearer regarding the "additional attributes". This mostly means pulling section 4 to sit between the introduction and section 2 in order to contextualize the "face" of the data presented in 2.
section 3 (video recording) needs further clarification on the methods and sampling design for the video footage, as well as its processing and analysis.
minor comments:
figure 1 is not necessary.
the reticula on the plotting area of figures 2, 3, 4, 6, 7, is not necessary. It is making the image more busy and could be removed to make them more readable;
observe that there are several references with error prompts throughout the text. please check them before producing the final version of the manuscript.
figure 5 is a correlation matrix. It needs more explanation on how the correlation was calculated to support it.
figures 6 and 7 could be merged into a single figure with A and B calls in the text.
table 1 is present twice in the manuscript. it should appear once and be called whenever necessary in text instead.
figures 10 and 13 are redundant, you can chose one of them and call it at both points on the text.
finally, the text could use a quick revision by someone fluent in English. there are a few mistakes as well as some mannerisms that are common on native romance language speakers.
Author Response
the manuscript presents a very robust and well collected set of data on nocturnal activity of sheep. the presentation, although not excellent, adequately contextualizes the data in the scientific state-of-the-art.
We thank the reviewer for his careful review of the paper, and for his valuable text editing suggestions.
However, the paper could benefit from some remodelling, especially in section 2. It should start with a comprehensive methodology for site selection, data collection and analysis/processing, and then follow-up with the data summary.
We edited according to your suggestion.
the 3-step presentation at the start of 2.2 informs very little of the steps, how they were performed, it mentions nothing on precautions and back-up plans for the case of data collection or transfer tools presenting problems, as well as not explaining how the "malformed" records are detected, why they are considered so, and needs to be clearer regarding the "additional attributes".
The 3-step presentation is explained immediately after the related figure, describing each step.
“A file contains records of different collars of a day.
This mostly means pulling section 4 to sit between the introduction and section 2 in order to contextualize the "face" of the data presented in 2.
We did it accordingto the suggestion.
section 3 (video recording) needs further clarification on the methods and sampling design for the video footage, as well as its processing and analysis.
Details were added to the text, and a paragraph was added to describe the data.
minor comments:
figure 1 is not necessary.
We remove the figure.
the reticula on the plotting area of figures 2, 3, 4, 6, 7, is not necessary. It is making the image more busy and could be removed to make them more readable;
The reticula was removed from the images.
observe that there are several references with error prompts throughout the text. please check them before producing the final version of the manuscript.
We fixed it.
figure 5 is a correlation matrix. It needs more explanation on how the correlation was calculated to support it.
We did it.
figures 6 and 7 could be merged into a single figure with A and B calls in the text.
We did it.
table 1 is present twice in the manuscript. it should appear once and be called whenever necessary in text instead.
We removed it.
figures 10 and 13 are redundant, you can chose one of them and call it at both points on the text.
We removed figure 10.
finally, the text could use a quick revision by someone fluent in English. there are a few mistakes as well as some mannerisms that are common on native romance language speakers.
We did it.
Reviewer 2 Report
The idea of measuring behaviour via collars is good and fits in well with current research approaches. However, if the data is to be used for behavioural analyses and especially night-time behaviour, as indicated, the data set needs to be fundamentally revised and supplemented.
The scientific question is not clear throughout the manuscript. Neither the abstract nor the introduction show a clear structure. It is mentioned that it is important to study behaviour in relation to feeding, dominance and reproduction. However, these behavioural effects cannot be investigated with the present data set, as corresponding behaviours were not recorded.
The title refers to nocturnal behaviour of sheep. Again, the dataset does not provide sufficient information. Sheep are mainly diurnal animals, so they rest and sleep at night. However, the behaviours lying, sleeping, resting were not recorded. Furthermore, the temporal distribution of these behaviours is extremely important for the behavioural interpretation of the data. With the available data set, the creation of a behavioural profile is only possible with gaps and only for individual animals.
Another problem arises from the temporal resolution of the data. Although it is stated that the collars were used 24/7, there are extremely different values in terms of records per weekday. The same applies to the records by status per hour. If there are large data gaps here, the interpretability of the data is not given as long as it is not clearly marked where the data gaps are. The presentation of the values in Fig. 3 is also unfortunate. Rarely occurring behaviours are not shown - shifting the scale would be helpful. Figs 1 and 2 are superfluous. The values can also be given in the text. In general, the quality of the Figures is inconsistent, poorly readable and unstructured.
The data set itself (excel file) has large gaps. Not all individuals are listed and a continuous recording of the animals over 10 seconds is not guaranteed without gaps. Individuals 2,4,9,11,13,14,16, 17,20 occur very regularly, whereas individuals 1,3,5,18, are practically absent from the data set.
Data validation is missing. How was the transmitter data evaluated? In order to evaluate the data set reliably, a comparison of the behaviours with the logger data and behavioural observation must be included. Information on how this was done in this study is not available. Video data is available for later comparison, but has not yet been analysed. However, it will be impossible to individually evaluate the behaviour of the animals in the videos and assign it to the corresponding logger data. This evaluation of the behaviours should therefore have been done by the researchers themselves.
The linking to the figures and tables is faulty (Error! Reference source not found) throughout the manuscript, which makes a stringent reading of the manuscript much more difficult. In Fig. 12 the source Google maps is missing.
Author Response
The idea of measuring behaviour via collars is good and fits in well with current research approaches. However, if the data is to be used for behavioural analyses and especially night-time behaviour, as indicated, the data set needs to be fundamentally revised and supplemented.
We thank the reviewer for his careful review of the paper, and for his comments.
The scientific question is not clear throughout the manuscript. Neither the abstract nor the introduction show a clear structure.
The rereading of the text after these weeks showed that both the abstract and the introduction were not written in the best way. We rewrote both and believe we have successfully resolved the issue.
It is mentioned that it is important to study behaviour in relation to feeding, dominance and reproduction. However, these behavioural effects cannot be investigated with the present data set, as corresponding behaviours were not recorded.
We agree and the text was edited.
The title refers to nocturnal behaviour of sheep. Again, the dataset does not provide sufficient information. Sheep are mainly diurnal animals, so they rest and sleep at night.
Title refers Sheep nocturnal activity that we monitored using iFarmTec collars. We agree that sheep behavedifferently during the night period, but at night they are not limited to resting and sleeping as the dataset shows. The videos confirm the nocturnal activity detected by the collar and allow us to see that the animals were walking, that they do not need light to move around and that they even eat. The analysis of the video (and the collar data) allows us to detect that there are always animals awake at night, which they do on a rotational basis, and according to what we have read, we believe that it has to do with a kind of survival instinct, with one of the sheep keeping watch over the flock.
However, the behaviours lying, sleeping, resting were not recorded. Furthermore, the temporal distribution of these behaviours is extremely important for the behavioural interpretation of the data.
In present work we focused on monitoring the activity, since we expected to measure quiet behavior with very few movements. This is the reason why we turned off the behavior classification the algorithm usually performs.
With the available data set, the creation of a behavioural profile is only possible with gaps and only for individual animals.
The dataset allows to create a profile of behavior dynamism, for each of the animals, since the records report to each of the collar. And collars were kept in the same animal throughout the complete study.
Another problem arises from the temporal resolution of the data. Although it is stated that the collars were used 24/7, there are extremely different values in terms of records per weekday.
It's true, . And after performing a preliminary analysis, it allowed to confirm the activity is consistent over the monitoring period.
The same applies to the records by status per hour. If there are large data gaps here, the interpretability of the data is not given as long as it is not clearly marked where the data gaps are.
It's true,but there may be an easy solution for researchers who want to use the dataset. If the objective is to have the same number of records on each day, it is possible to simply remove the data collected during the day period when the animals were left in confinement. In this case, we preferred to not remove information that may be useful in the future.
The presentation of the values in Fig. 3 is also unfortunate. Rarely occurring behaviours are not shown - shifting the scale would be helpful.
Yes, it is true and we represented it in a logarithmic scale.
Figs 1 and 2 are superfluous. The values can also be given in the text. In general, the quality of the Figures is inconsistent, poorly readable and unstructured.
We removed it and we improved figures quality.
The data set itself (excel file) has large gaps. Not all individuals are listed and a continuous recording of the animals over 10 seconds is not guaranteed without gaps. Individuals 2,4,9,11,13,14,16, 17,20 occur very regularly, whereas individuals 1,3,5,18, are practically absent from the data set.
Yes, it is true. In fact, we relied on a plan to charge the batteries that was not flawlessly performed and we were very disappointed as well when those gaps where discovered. But the dataset maintains its value since it represents the dynamic behavior of the animals during the night period, with a very nice sampling rate during a considerable time period. It allows to study the animals’ activity and compare them.
Data validation is missing. How was the transmitter data evaluated? In order to evaluate the data set reliably, a comparison of the behaviours with the logger data and behavioural observation must be included. Information on how this was done in this study is not available.
iFarmTec collar was developed based on a supervised machine learning process, and the decision tree learning accuracy compares very well with commercial inertial data loggers. Just as a curiosity, details on the development of iFarmTec collar are available in references [17][18][22] and [36].
There is no information on the states collar classifies, except the references to the papers published during the development phase, because our goal was to access animal activity during the night period. Thus, we decided to undervalue the behavior classification provided by the collar data. Moreover, ML process that enabled the classification algorithm was performed in a pasture context, not during the night period. As a result, a lower accuracy in the current scenario might be expected.
Video data is available for later comparison, but has not yet been analysed. However, it will be impossible to individually evaluate the behaviour of the animals in the videos and assign it to the corresponding logger data. This evaluation of the behaviours should therefore have been done by the researchers themselves.
We agree that the task of identifying which animal is moving at each moment is laborious, as this depends on the sequential analysis of the video images. This task may not actually be necessary, as the sensor data (distance from neck to ground, accelerations) is self-contained. It allows to analyze their activity and it allows to take conclusions on the their sleeping or lying habits. There is for instance a study about goats’ lying habits where the authors quantify those periods and refer whether the animals were liyng to the left or the right side (refª ????).
As a curiosity the video was not planned at the beginning of the research. We started to video record the animals since data showed an unexpected activity of animals moving throughout the night, in a very quiet place where disturbances were not expected. Video confirmed the sheep walked around during the night, in the dark, without stepping on their mates. . Additionaly, they ate.
Moreover, the journal's instructions in the "Data Description" papers are to describe the data, to describe the capture process, but not to analyze or draw conclusions on the data.
The linking to the figures and tables is faulty (Error! Reference source not found) throughout the manuscript, which makes a stringent reading of the manuscript much more difficult.
It was an unforgivable mistake on our part that had to do with the exchange of the submitted pdf file. It was fixed.
In Fig. 12 the source Google maps is missing.
Yes it somehow disappeared, but we added it again.
Round 2
Reviewer 2 Report
Thank you for the revision of the manuscript. The dataset provides an important contribution to better understand the nocturnal behaviour of animals. In order to be able to make more detailed analyses of behaviour, future studies should urgently take care to also include resting/sleeping behaviour.
Abstracts and introduction: Both are clearer now. However, the last sentence of the abstract needs to be changed: "With the collection of data from various sensors, we can identify their resting and sleeping activity, throughout the night, and to count and measure the average time of rest periods. The data set does not give any information about resting or sleeping activity.
Figure 5: The order of the days is still confusing. Please put them in the right order.
Author Response
We thank the reviewer for his comments and suggestions. We edited the document and we hope we successfully addressed reviewer suggestions.
We agree the dataset does not allow to differentiate between sleep and rest activities, and we corrected the sentence accordingly.
We followed reviewer suggestion as well and we edited Fig. 5.
Author Response File: 
Author Response.pdf
