Comparing Body Density of Lumpfish (Cyclopterus lumpus) to Different Operational Welfare Indicators

Round 1

Reviewer 1 Report

General comments:

This is an interesting paper that aims to compare the differences in body density, OWIs, and other indicators between lumpfish from five locations. The relationships between body density and other indicators were also detected. The logic of this study is simple and fluent, and my major concern focuses on the statistical methods (see specific comments). I think that the present results may be not robust, and the results may have sharp changes after altering the statistical models, so I do not review the Discussion section in this review round.

Specific comments:

L10: Please add the scientific name to the lumpfish.

L110: I do not understand how to get the volume of fish (v). Please elaborate on this point.

L144-146: What are your independent variables? Were the normality and heterogeneity of data detected before ANOVAs? Please add more information about these points. In addition, I do not think it is proper to apply ANOVA to deal with your data. Particularly, the liver color, skin score and fin score are categorical data, and their expected distribution types are not normality, but may be Possion (or other types) distribution. Unfortunately, ANOVA can not deal with the such data type. The proper model, I think, for dealing with these data is generalized linear mixed model (GLMM). I strongly suggest the authors consider this point.

L146-148: This method is odd to me. Why did not you choose the linear mixed model (LMM) or GLMM to detect the correlation between body density and OWIs? I think LMM or GLMM is more robust than the present model. In the LMM or GLMM, you can include sampling site and sampling date (and other factors that you think maybe affect statistical outcomes) as random factors, use OWIs as independent variables, and use body density as a dependent variable. The LMM and GLMM can be constructed in R software. The choice of using LMM or using GLMM depends on the distribution type of your data: for continuous data, LMM is proper, but for categorical data, GLMM is more proper.

L154-155: Was not the body density affected by disease burst? Why did the authors include the social density of location D in their analysis? Please elaborate on this point. I think that it may be more proper to entirely delete the data from location D. I am expectant to hear the author’s opinion.

L164-165: Please add information on sampling size and type of error bar (i.e., SD or SE) to the figure caption.

L169: It is very surprising for me that, the author detected a significant correlation between Fulton’s K and body density, under the circumstance that R square was very low (0.008). I speculate that the reason for this abnormal result may be related to the improper model that the authors applied. Please elaborate on this point.

L188-189: Please describe this significant result in detail. The stomach fulness is an important indicator for fish welfare, especially in practical aquaculture.

L198: There seems a miswriting (‘HIS’).

L212: There seems a miswriting in the x-axis (‘length’).

Author Response

Reviewer 1

This is an interesting paper that aims to compare the differences in body density, OWIs, and other indicators between lumpfish from five locations. The relationships between body density and other indicators were also detected. The logic of this study is simple and fluent, and my major concern focuses on the statistical methods (see specific comments). I think that the present results may be not robust, and the results may have sharp changes after altering the statistical models, so I do not review the Discussion section in this review round.

Specific comments:

L10: Please add the scientific name to the lumpfish.

• Changed as suggested.

L110: I do not understand how to get the volume of fish (v). Please elaborate on this point.

• All calculations in Eq.2.1-2.4. To calculate the volume of the fish we first weighted the fish in water and air thereby getting the upthrust from the water (u).
• Sea water density (r) was calculated according to the set of formulas given in Eq. 2.4 (note that the journal gives the wrong symbol, should be r not Ρ).
• The volume (v) of the fish (in ml) was defined as: v = u/ ρ (i.e. upthrust/sea water density).

L144-146: What are your independent variables?

• The independent variables are the locations (Figs. 1, 3 and 5, locations A-E) or the liver colour (Fig. 4, liver colour 1-5). This info has been added to the ms.

Were the normality and heterogeneity of data detected before ANOVAs? Please add more information about these points.

• Yes, we assess normality of distributions with a Kolmogorov-Smirnov test, and homogeneity of variances was tested using Levene’s F test. This info has been added to the revised version of the ms.

In addition, I do not think it is proper to apply ANOVA to deal with your data. Particularly, the liver color, skin score and fin score are categorical data, and their expected distribution types are not normality, but may be Possion (or other types) distribution. Unfortunately, ANOVA can not deal with the such data type. The proper model, I think, for dealing with these data is generalized linear mixed model (GLMM). I strongly suggest the authors consider this point.

• Yes, it is correctly noted by the reviewer that colour, skin, fin and stomach score are categorical data.
• However, the dependent variable in all cases is the fish density which is a typical continuous variable justifying the use of regular GLM methods (here ANOVA and ANCOVA). This is valid for our comparison of body density at different locations (Figs. 1-2) and of different liver colours (Fig. 4).
• But we agree that for the fin, skin and stomach score using ANOVA is not strictly correct. However, these data are only used for descriptive purposes.
• We have also included a description of the statistical method used for Figs. 3 and 6-8 as this was lacking in the previous version of the ms.

L146-148: This method is odd to me. Why did not you choose the linear mixed model (LMM) or GLMM to detect the correlation between body density and OWIs? I think LMM or GLMM is more robust than the present model. In the LMM or GLMM, you can include sampling site and sampling date (and other factors that you think maybe affect statistical outcomes) as random factors, use OWIs as independent variables, and use body density as a dependent variable. The LMM and GLMM can be constructed in R software. The choice of using LMM or using GLMM depends on the distribution type of your data: for continuous data, LMM is proper, but for categorical data, GLMM is more proper.

• Thank you for noting this. By mistake we described the wrong method (so understandable it seemed odd to the reviewer as it was wrong in this case) used for the possible correlation between fish body density and Fulton´s K, HSI, weight and length. As no linear dependence was assumed we investigated possible correlation using the Pearson product-moment correlation coefficient.
• Unfortunately, we don’t have sufficient data to break the analysis down by sampling site or sampling date so in this case we have focussed on the overall trend between the variables.

L154-155: Was not the body density affected by disease burst? Why did the authors include the body density of location D in their analysis? Please elaborate on this point. I think that it may be more proper to entirely delete the data from location D. I am expectant to hear the author’s opinion.

• We don’t really know if the fish body density was affected or not, but our data do not point to much effect on the density (see Figs. 1-2).
• However, we noted change in the liver colour in fish from location D and have therefore omitted this in most of the comparison i.e., when looking at relationship between density and measured biological variables (liver colour, Fulton, HSI, length and weight).

L164-165: Please add information on sampling size and type of error bar (i.e., SD or SE) to the figure caption.

• This information has been added.

L169: It is very surprising for me that, the author detected a significant correlation between Fulton’s K and body density, under the circumstance that R square was very low (0.008). I speculate that the reason for this abnormal result may be related to the improper model that the authors applied. Please elaborate on this point.

• There is a large variation in this data set, so the trend is rather vague. As pointed out earlier the relationship was tested with a simple correlation test (Pearson r) as no dependency was assumed a priori between the two variables.
• However, it seems logical that these two variables are negatively correlated as a fish with a high Fulton´s K is fatter and most likely has a larger lipid reserve, which is one of the methods of buoyancy in lumpfish. This is discussed on page 11 in the ms.

L188-189: Please describe this significant result in detail. The stomach fulness is an important indicator for fish welfare, especially in practical aquaculture.

• We have added a more thorough explanation of this result as stomach fulness is important as noted by the reviewer. This is also discussed in more details on page 10 (lines 250-256).

L198: There seems a miswriting (‘HIS’).

• Yes, and this has been corrected.

L212: There seems a miswriting in the x-axis (‘length’).

• This has been corrected.

Author Response File: Author Response.docx

Reviewer 2 Report

-   Dear authors,

thank you for conducting this study. There are a few comments which may improve the manuscript a bit.

For Fig 1 and Fig 2 and Fig 5: I would prefer to see a boxplot and not a bar plot. A boxplot more nicely shows the range of the data.

-For Fig 3 and Fig 6 and Fig 7, was a correlation calculated. Then there should also be a p value for that and a correlation value.

 Why was there no correlation between the density calculations and the weight of the fish. I would have expected that since the weight was used to calculate the body density.

Would it make sense to calculate an overall welfare index from the OWI and separately for OWI plus the body density and perform a classification analysis thereafter in order to see if misclassification increases or decreases after the addition of body density as a variable?

Author Response

Reviewer 2

Dear authors,

thank you for conducting this study. There are a few comments which may improve the manuscript a bit.

For Fig 1 and Fig 2 and Fig 5: I would prefer to see a boxplot and not a bar plot. A boxplot more nicely shows the range of the data.

• We understand the reviewers comments on this topic, but the range of the data is already shown through the SD displayed on the figures so we don’t see the need to change to a boxplot.

-For Fig 3 and Fig 6 and Fig 7, was a correlation calculated. Then there should also be a p value for that and a correlation value.

- Yes, this data was tested with as simple correlation and the result has been added to the text.

 Why was there no correlation between the density calculations and the weight of the fish. I would have expected that since the weight was used to calculate the body density.

• Yes, this was a surprising result for us too.
• We suspect that the large variation in the data "blurred" for a statistical correlation on these data. When compared Figures 7 and 8 show similar trends so the size of the fish was, overall, found the be correlated to the body density as one would expect.

Would it make sense to calculate an overall welfare index from the OWI and separately for OWI plus the body density and perform a classification analysis thereafter in order to see if misclassification increases or decreases after the addition of body density as a variable?

• Yes, this is good suggestion. But unfortunately, we only scored the fin, skin and stomach filling so it is not possible to make on overall OWI score for the fish.
• However, fish with the highest density tended (but not significantly) to have lower fin and skin OWI score, but also lower stomach filling (as several of those investigated had empty stomachs and hence scored 1).

Author Response File: Author Response.docx