Non-Linear Analyses of Fish Behaviours in Response to Aquatic Environmental Pollutants—A Review

Round 1

Reviewer 1 Report

This paper gives a comprehensive review about analyses of fish behavioural responses to environmental pollutants. This topic is very interesting and the conclusion can be helpful to the modern fish aquaculture.  I have a few commetns:

1. Try to avoid words such as "Possibly" in a scientific publication.

2.  The authors reviewed many pollutants effects to fish. It would be better to add tables in some of these sections. It will be more intuitive to show the relationship between fish physiological variables or behavior and the pollutants type.

Author Response

Answers to Reviewer 1, Round 1

Comments and Suggestions for Authors

This paper gives a comprehensive review about analyses of fish behavioural responses to environmental pollutants. This topic is very interesting and the conclusion can be helpful to the modern fish aquaculture.  I have a few comments:

A: We thank the reviewer for the positive comments. It follows the answer to each comment indicating the modifications that have been made in the revised version.

Q1. Try to avoid words such as "Possibly" in a scientific publication.

A1: There were two “possibly” in the original manuscript, one in line 294: “Possibly the most complete work…” that has been substituted in line 294 of the revised version, by “In our opinion, the most complete work….”. The second one, in line 570: “…considering the work by Eguiraun et al. [48] possibly most of the fish in these published works were subjected…” has been substituted (see lines 570-571 of the revised version) by “… based on the results by Eguiraun et al. [48], most of the fish used in these works were probably …”

Q2.  The authors reviewed many pollutants effects to fish. It would be better to add tables. physiological variables or behavior and the pollutants type.

A2: We agree with the reviewer, that is why there are 2 tables at the end on the manuscript summarizing the results. Table A1 orders the reviewed works by pollutant type, which we believe is what the referee means. Unfortunately, we cannot give the “physiological variables” because many of these works do not perform biochemical or physiological analyses. The changes in the behaviours are given in Table A2. We prefer to present all the works in only 2 tables, rather than prepare a table for each contaminant, for convenience, i.e., so that the reader has a general overview of all the works at a glance.

Reviewer 2 Report

The authors reviews technical papers for the analysis of fish behavior which is an effective way to indirectly identify the health of fish.  However, this paper should be modified due to the following concerns.

1. The authors claims that fractal dimension (FD) and entropy are two groups of such non-linear analyzing methods that serve as indicators of the complexity (FD) and predictability (Entropy) of the behaviours. However, the semantic gap between signal processing results of the sensor data and fish behaviors are often unclear. The authors should first discuss how to make sure the conclusion is correct when we use signal processing tools to predict fish healthy conditions using some experimental results.

2. Why we can just analyze the fractal dimension (FD) and entropy for predicting fish behaviors? In the usage of non-linear analysis, we can find many signal processing tools. Is it possible to have a complete review to tell the reader how to use non-linear signal processing tools for fish healthy prediction?

3. The signals captured fro the fish which is lively swimming in the underwater environments are often time related. Why time-series data analytics are not reviewed to clarify the relationship between non-linear signal analytics and fish behaviors?

4. Does there exist any methodology for fish behaviors using non-linear signal processing? Please include the reviews of the methodology in the revised manuscript. 

Author Response

Answers to Reviewer 2, Round 1

Comments and Suggestions for Authors

We thank the reviewer for the interesting comments. It follows the answer to each comment indicating the modifications that have been made in the revised version.

Q1: The authors reviews technical papers for the analysis of fish behavior which is an effective way to indirectly identify the health of fish.  However, this paper should be modified due to the following concerns.

A1: Our review does not deal with the analysis of fish behaviour to indirectly identify its health. It compiles works showing that contaminants may alter two properties of fish behaviours: their predictability and their complexity. This alteration is important because it can take place prior to the development of diseases or death. Once the disease in set, the behavioural studies and early warning loses its application, and the farmer should make use of traditional veterinary treatments or culling of the production. We have explicitly indicated this in lines 472-481: “The agreement in the trends between the non-linear parameters and the biochemical analysis before the results become significant, and thus significantly affect the health of the fish, confirms the value of the non-linear assessments as warning systems of negative events (toxicity in this case) affecting the fish in the early stages. It is important to stress that none of the results were statistically significant and most likely an external observer would have disregarded them for that reason concluding that the system was working fine. However, we do know that the fish was being intoxicated. Consequently, we wish to emphasize the relevance of considering the trends of evolution of the parameters earlier in their development rather than waiting until the values become statistically significant, when the harm may be irreparable”. Indeed, the relevance of these analyses lays on their capability to detect alterations in behaviour before the animal´s health is compromised, and to set up the means to avoid further deterioration of the system. Furthermore, alterations in the system do not necessarily indicate that the fish is unhealthy, it may just be temporarily stressed for example, but, if not corrected on time, it may end up with more serious consequences, including diseases. As mentioned, once the animal is sick, one should use standard diagnostic methods to identify the causes and treat it.

Q2. The authors claims that fractal dimension (FD) and entropy are two groups of such non-linear analyzing methods that serve as indicators of the complexity (FD) and predictability (Entropy) of the behaviours. However, the semantic gap between signal processing results of the sensor data and fish behaviors are often unclear.

A2: We present a review paper and as such the claim that FD and entropy are two types of non-linear analyses indicating the complexity and predictability of biological systems (including, animals’ behaviour) is not our own personal claim; rather, it summarizes the claims of numerous authors, based on their scientific experimental works as mentioned in the manuscript: see section “1.3. Fractal dimension (FD) and Entropy properties in biological systems” of the original version, lines 219-251, and references 54, 55 and 64-76. Therefore, “the semantic gap between signal processing results of the sensor data and fish behaviors” are dealt with in the individual scientific research papers cited. We are aware of the numerous methods available for signal processing and of the variety of tools to obtain signals (noise, images, etc.) but a discussion on that technical issue is not the subject of the present review; as clearly indicated in lines 267-271 of the last paragraph of section 1.4 Aim of the work: “The purposes of this work are i) to give an overview of recent works reporting the effects of relevant environmental contaminants on the FD and Entropy on individual and collective behavioural responses of fish, and ii) to indicate their potential applications to identify the presence of undesirable environmental contaminants in either fish farming settings or natural populations. “

Q3: The authors should first discuss how to make sure the conclusion is correct when we use signal processing tools to predict fish healthy conditions using some experimental results.

A3: As mentioned above, neither the review not the specific research articles cited aim at “predicting fish health condition”, the review is limited to detecting alterations in their behaviour caused by chemical substances. The review does not target the analysis in detail of the specifics of the materials and methods of each paper to reach the final conclusions. All the cited papers must have successfully gone through that particular scrutiny prior to being accepted for publication, since all the papers used have been published in scientifically refereed international highly reputed journals. Moreover, none of the papers cited indicate that a specific health condition can be predicted by these nonlinear behavioural analyses, since they are not specific for any particular health condition. For prediction purposes there is a set of different mathematical treatment of data that use classification, deep learning and AI techniques, but that is not the subject of our review. As mentioned in answer A2, the aim of the work is clearly stated in lines 267-271, last paragraph of section 1.4 Aim of the Work. The relevance of using the nonlinear behavioural analyses on which our review focuses, lays in their ability to detect changes in their values or in their trends that indicate changes in the biological system, they are not disease-diagnostic tools. This is indicated in the section of General Conclusions, lines 602-607 of both the original and revised versions: “Indeed, the response of fish systems to the simultaneous exposure to more than one stressor is neither synergistic nor linear, but complex and non-lineal [42,48,53,88]. In addition, the use of FD or entropy to aim at identifying particular contaminants is not recommended, as has been previously stressed [42,43]. Thus, alterations in FD or entropy must never be considered indicators of a given contaminant, only of the fact that something unexpected is affecting the fish.”

Q4. Why we can just analyze the fractal dimension (FD) and entropy for predicting fish behaviors?

A4: We cannot. Nowhere in the text it is indicated that FD or entropy can be used to predict fish behaviours. They are used to analyze actual/real fish behaviours and indicate whether the results are within normal values (i.e. the values that are usually calculated when the systems is working normally) or outside the range of normal values (and then the farmer must start investigating what has caused the difference in the predictability and/or complexity of the system). It must be stressed that the entropy of a system (an indicator of its energy or unpredictability) is literally an indicator of its predictability (i.e., the predictability/unpredictability of the given behaviour used to calculate the entropy) and it is not used as a predictor for a given condition (disease, intoxication, stress etc.). Predictability and prediction are not the same: the predictability, calculated according to the systems entropy, is a number, it can be high or low, and the prediction is a probability: probability of the fish being sick given a set of conditions and it is a percentage. To predict something (behaviours or diseases) one would need to use complex classification and prediction techniques as mentioned above in answer A3: “For prediction purposes there is a set of different mathematical treatment of data that use classification, deep learning and AI techniques, but that is not the subject of our review”.

Q5: In the usage of non-linear analysis, we can find many signal processing tools. Is it possible to have a complete review….?

A5: The reviewer is correct, there are many signal processing tools, as well as many different types of equipment for signal-acquisition. To review them will indeed require a new complete review on a subject that has been already been targeted by other authors. We needed to set up limits to the present review. As mentioned in answers A2 and A3, the boundaries of the current review are clearly indicated in lines 267-271 of the last paragraph of section 1.4 Aim of the work.

Q6: Is it possible to have a complete review to tell the reader how to use non-linear signal processing tools for fish healthy prediction?

A6: As already mentioned several times before, the review does not deal with predictions. The issue of prediction (health, behaviour etc.) depends on the availability of reliable databases for both, control cases and for each targeted case (each disease, each behaviour, each species and so on) together with the application of selected complex and adequate data treatment and classification algorithms (AI techniques) to reach reliable predictions. The purpose of the present review is not to study data classification and prediction algorithms. As mentioned in answers A2, A3 and A5 the boundaries of the current review are clearly indicated in lines 267-271 of the last paragraph of section 1.4 Aim of the work. In any case, the absence of reliable databases makes the implementation of the classification and prediction techniques too unreliable. Also, this review does not target fish health, just the presence of contaminants. Therefore, the reviewer’s request to somehow introduce an extra, completely different review on fish health prediction is not possible in the current context.

Q7: The signals captured from the fish which is lively swimming in the underwater environments are often time related. Why time-series data analytics are not reviewed to clarify the relationship between non-linear signal analytics and fish behaviors?

A7: We have not reviewed the usefulness of time-series data because all the data obtained by the cited papers are indeed obtained during periods of time, i.e. they all analyze time-series of data. The quality of the data treatment is guaranteed by the scientific and critical revision each paper has gone through prior to publication. If the reviewer refers to differences in the natural behaviour of the fish due to time of the day, season of the year, sex, aggression etc, there are indeed other published works dealing with those factors, some of which are cited in this review, but, again, that is not the focus of the current review. If the reviewer refers to a review on the different data analysis techniques for time-series data from a mathematical point of view, that is also outside the scope of our work and would be more suitable for a more specialized mathematical/engineering journal.

Q8. Does there exist any methodology for fish behaviors using non-linear signal processing? Please include the reviews of the methodology in the revised manuscript 

A8: There are indeed methodologies for fish behaviour analysis using non-linear signal processing and the manuscript reviews them. As clearly stated by the title of the work: “Non-linear Analyses of Fish Behaviours in Response to Aquatic Pollutants. A review” this entire review is devoted to them, and the two targeted techniques are the fractal dimension and the entropy of the behaviours. Thus, the reviews of the methodology are included in the revised manuscript. Accordingly, the original and revised manuscript already satisfy this request. If, on the other hand, the reviewer refers to the existence of additional nonlinear signal processing techniques (i.e., in addition to FD and entropy) applied to fish behavioural studies, a literature search in Scopus (dated on the 02/06/2023) with the string  (TITLE-ABS-KEY("fish behavio*") AND TITLE-ABS-KEY ("non-linear analys*" OR “nonlinear analys*”) AND NOT TITLE-ABS-KEY (entropy) AND NOT TITLE-ABS-KEY("fractal dimension")) has rendered zero hits. Therefore, as far as we known, the original manuscript already reviews all the published nonlinear methodologies available for fish behavioural analysis.

Round 2

Reviewer 2 Report

Although I think the authors do not any siginicant changes on the paper, no further comments can be given.