Depth of Edge Influence in a Madagascar Lowland Rainforest and Its Effects on Lemurs’ Abundance

Round 1

Reviewer 1 Report

The authors selected 9 sample zones in the lowland rainforest area of Madagascar to establish long-term monitoring sample plots, and studied the influence of edge effect on the number of lemurs in this area. The research can provide theoretical basis for formulating fragmented forest protection strategies. I think this study has important implications, but some details need to be corrected. The specific problems are as follows:

 

1.[Line134-135] Figure 1. The meaning of the picture is unclear. I couldn't see the nine transect zones, the location of the rainforests in the study area, geographic information, etc.

 

2. It is suggested that the edge effect and area effect can be more clearly distinguished and the interference between them can be reduced in the future study of lemur edge effect; In many cases, area greatly affects the number of species.

 

3. However, studies have proved that human activities have a great impact on the number of species, and this interference is far-reaching, and the vegetation status, climate conditions, and human activities in the study area will have direct or indirect effects on the number of species, which are not discussed in the paper.

 

 

4. The references and grammar of the article need to be further polished to make it easier to understand

Author Response

The authors selected 9 sample zones in the lowland rainforest area of Madagascar to establish long-term monitoring sample plots, and studied the influence of edge effect on the number of lemurs in this area. The research can provide theoretical basis for formulating fragmented forest protection strategies. I think this study has important implications, but some details need to be corrected. The specific problems are as follows:

We thank the reviewer for acknowledging the importance of our study and for providing very useful suggestions.

1.[Line134-135] Figure 1. The meaning of the picture is unclear. I couldn't see the nine transect zones, the location of the rainforests in the study area, geographic information, etc.

We created a new map also based on the comments from the other reviewer. We hope that now the transects and sub-transects are clear

2. It is suggested that the edge effect and area effect can be more clearly distinguished and the interference between them can be reduced in the future study of lemur edge effect; In many cases, area greatly affects the number of species.

We acknowledge that there is a relationship between species number and forest area and that as the latter decreases in size the former, ultimately, has to decrease as well. However, our study site is part of a continuous lowland rainforest, the Tsitongambarika Protected area, of 605 km², so we cannot really consider area size in our investigation (we are not working in different fragments). 

3. However, studies have proved that human activities have a great impact on the number of species, and this interference is far-reaching, and the vegetation status, climate conditions, and human activities in the study area will have direct or indirect effects on the number of species, which are not discussed in the paper.

We totally agree, and we now have included a more general statement at the beginning of page 2 that humans, through direct and indirect interference, have an effect on what species can survive where. A statement that other factors, including human activities, can impact abundance is also present in the discussion (section 4.3).

4. The references and grammar of the article need to be further polished to make it easier to understand

We have carefully edited the paper to ensure we are as clear in our language as we can be, and we have ensured that the spelling and grammar are consistent and correct.

Reviewer 2 Report

Dear Authors,

 

              This is a very interesting study about edge effect on lemurs’ abundance in a Madagascar lowland rainforest. These ecotone areas could represent an important habitat structure for some animal species and these requires specific studies in order to quantify their value. On the other side, the forest edges will get a higher percent in the forest structure due to many clear cuts and conversion of the forest into agriculture areas. So, is highly important to understand the ecology of this particular habitat structure and also the species reaction to be able to adopt suitable conservation measures, if we want to protect our plant and animal species.

              I have some general comments regarding each section of the article and also some minor comments where I specified the lines from the manuscript.

              In the introduction you mention that your study is focused on DEI determination in TGK, identifying the response of six lemur species and estimate the density of lemur abundance, but in methods you speak first about lemurs and after about vegetation. Which is the correct order? I consider that lemur first and after vegetation because your study is focused mainly on these species, but you have to decide.

For Methods section you should explain better the survey design. How many transects and sub-transects?; how did you chose them?; how did you divided them? How long they are? How did you choose the vegetation sampling plots? Is there any random or stratified process for transect selection? This section should be improved in order to be clear for your readers!

              Did you used Generalised Linear Mixt Models or General Linear Model? The glmmTMB package is for GLMM. This test is not well explained. Which are the response factors, which are the predicted variables and which are the random factors? In results section I do not see the results of this test! Did you check the multicollinearity?

              The first letter of the animals and plants common names is always capitalized.

              The Figure 1 (map) should illustrate the land use categories for the study site in order to observe the habitat matrix. You should add also the sub-transects of this map!

For the Results section you need a table with species presented for each surveyed month will be more suitable. Did you record each species every month? Can you make also a graphic with their presence in your transects per month? Did you consider month as a random factor for your analysis? Also, it will be interesting to see the results of day and night transects.

You have to specify the results of GLMM or GLM test, in a separate table, and also you have to specify how did you chose the best model?

Can you present the results of the vegetation survey? These are very important in order to check the forest characteristics!

 

 

Lines 48 – 49: These percent are for forest or tropical forest at global scale? I am asking these because at lines 80-83 are different percent for the Madagascar forest.

Lines 51 – 54: Here are all the possibilities for the edge responses. You can delete this, of if you want to keep it, please give some examples for each response category.

Lines 80 – 83: Please check my comment regarding Lines 48 – 49.

Lines 101 – 106: The first letter of the common names is always capitalized.

Lines 101 – 113 are more suitable for discussion section because you must say why you predict these, so you have to argue this prediction, you should explain table 1!

Line 118: You mentioned in line 21 that the study was conducted from May 2015 to July 2016, which is the correct date?

Lines 121 – 122: The parenthesis should be after the maximum altitude.

Line 131: Please check my comment on line 118.

Line 136 – 137: On the map we can not understand which are those 9 transects. I suppose that in 3 cases two transect are connected. Can you mark them?

Lines 157 – 170: Can you give in Supplementary materials the species list? These will help us to understand the species composition of these forests.

Lines 177 – 183: You had 9 transects (in the map caption) how you dived them to get 11 edge sub-transects and 10 interior sub-transects? How long are these transects and sub-transects? Are they at the same length?

Lines 183 – 194: Did you check the models for multicollinearity? You can use Variance Inflation Factor (VIF)

Lines 200 – 207: Can you add here a table with your distance models and their AICc score? To present your models in order to show also the differences between the distance models.

Lines 210 – 212: Can you give the test which was used to find out if there was a significant variation or no? Also, the results of the test?

Line 215: you can not cite a table in the Figure caption. You cand add this explanation in the Results text. You have 33 or 32 vegetation plots?

Line 219: You have 33 or 32 vegetation plots?

Lines 222 – 226 can you give also the values of the comparation test? Also, p values, especially for those significantly higher?

Lines 241 – 244: What is the difference between Density of groups and Density of individuals?

Results section: I do not see the GLM results! Have you used this test, as you mention in Methods section?

Lines 251 – 253: Do not repeat the information, just clarify these percent according to my previous comments!

Lines 332 – 378: You should discuss each lemur species. Compare their density in the study site with other areas and studies and explain why there are differences.

Discussion: You can add also some information about species ecology, their preference for different habitats, food, hibernation and others which are suitable to be presented according to your study.

 

Best regards

Author Response

Dear Authors,

              This is a very interesting study about edge effect on lemurs’ abundance in a Madagascar lowland rainforest. These ecotone areas could represent an important habitat structure for some animal species and these requires specific studies in order to quantify their value. On the other side, the forest edges will get a higher percent in the forest structure due to many clear cuts and conversion of the forest into agriculture areas. So, is highly important to understand the ecology of this particular habitat structure and also the species reaction to be able to adopt suitable conservation measures, if we want to protect our plant and animal species.

We thank the reviewer for acknowledging the importance of our study and for providing very useful suggestions.

              I have some general comments regarding each section of the article and also some minor comments where I specified the lines from the manuscript.

              In the introduction you mention that your study is focused on DEI determination in TGK, identifying the response of six lemur species and estimate the density of lemur abundance, but in methods you speak first about lemurs and after about vegetation. Which is the correct order? I consider that lemur first and after vegetation because your study is focused mainly on these species, but you have to decide.

We would like to keep vegetation first simply because we based the selection on edge vs interior based on the vegetation. So vegetation results in our view should come first. We have now edited the methods accordingly.

For Methods section you should explain better the survey design. How many transects and sub-transects?; how did you chose them?; how did you divided them? How long they are? How did you choose the vegetation sampling plots? Is there any random or stratified process for transect selection? This section should be improved in order to be clear for your readers!

We understand that the information might have been in different sections, and other information were missing. We have now restructured the methods so that the survey design should be more clear.

 

             Did you used Generalised Linear Mixt Models or General Linear Model? The glmmTMB package is for GLMM. This test is not well explained. Which are the response factors, which are the predicted variables and which are the random factors? In results section I do not see the results of this test! Did you check the multicollinearity?

We have now specified in the data analysis section, but mainly the encounter rates of individuals and groups are the response variables and the transect type (edge vs interior) is the factor. No other factors are present as our aim was only to understand differences in edge vs interior. No multicollinearity check is done (only one factor). glmmTMB can do GLM or GLMM but the main characteristics of this package is to handle count data, proportional data, and zero-inflated data.  

              The first letter of the animals and plants common names is always capitalized.

I am afraid we have to disagree on this one. This rule might be true for some disciplines, e.g., ornithology, but for most of fauna and flora the rule is to have common name lower case.

              The Figure 1 (map) should illustrate the land use categories for the study site in order to observe the habitat matrix. You should add also the sub-transects of this map!

We created a new map also based on the comments from the other reviewer. We hope that now the transects and sub-transects are clear

For the Results section you need a table with species presented for each surveyed month will be more suitable. Did you record each species every month? Can you make also a graphic with their presence in your transects per month? Did you consider month as a random factor for your analysis? Also, it will be interesting to see the results of day and night transects.

Exploring seasonal variation was not an aim of our study, that is why we did not include monthly figures. For dwarf lemur we excluded the months when the species is hibernating. The other lemur species were observed every month .  What know that the lemurs can modify their behaviour, diet, home range size, and daily path lengths seasonally but again it is not the focus of this paper to test how those variations affect the densities . Also a separation day vs night in our view is not needed  as brown lemur and bamboo lemur were detected only during day transects and the other lemurs were only detected during night transects. This information is in section 2.4.

You have to specify the results of GLMM or GLM test, in a separate table, and also you have to specify how did you chose the best model?

We have now added a table with the full results. We did not include that in the previous version of the draft since we only have one factor, so estimated marginal means and p-value are enough to see the differences, but we fully agree that the full results should be added (to the appendix as we are hesitant to break the narrative with too much information). We specified that for the model selection we tested different families included in glmmTMB and selected Tweedie since it was the only one that passes all the tests for model fit based on QQ plot residuals and residual vs predicted plot from the package “DHARMa”.

Can you present the results of the vegetation survey? These are very important in order to check the forest characteristics!

We agree that some readers would like to see that information and we added the species list as suggested. We also added the presence in edge vs interior. We added this information to the appendix with a brief explanation in the results.

Lines 48 – 49: These percent are for forest or tropical forest at global scale? I am asking these because at lines 80-83 are different percent for the Madagascar forest.

Yes these are global estimates, at that point of the introduction we stayed broad. We have now specified that.

Lines 51 – 54: Here are all the possibilities for the edge responses. You can delete this, of if you want to keep it, please give some examples for each response category.

We think that it is important to keep the three different responses as those are the basis for our predictions. We did include some examples in the following paragraph.

Lines 80 – 83: Please check my comment regarding Lines 48 – 49.

Changed

Lines 101 – 106: The first letter of the common names is always capitalized.

Again we have to disagree on this one, see above. However, we realise Southern was lower case in some sections and we edited that

Lines 101 – 113 are more suitable for discussion section because you must say why you predict these, so you have to argue this prediction, you should explain table 1!

We based our predictions on the information provided in the introduction (lines 56-67). We think it is important to present them first in the introduction and discuss them later in the discussion.

Line 118: You mentioned in line 21 that the study was conducted from May 2015 to July 2016, which is the correct date?

May 15 to July 16 is correct but we only have rainfall data from July 15. Now edited that sentence

Lines 121 – 122: The parenthesis should be after the maximum altitude.

Thanks. The parenthesis refers to the lowland rainforest not the maximum altitude

Line 131: Please check my comment on line 118.

Checked. Thanks.

Line 136 – 137: On the map we can not understand which are those 9 transects. I suppose that in 3 cases two transect are connected. Can you mark them?

We have now changed the map. The trails are connected but transects are not connected, we have now edited that. We hope this is now more clear.

Lines 157 – 170: Can you give in Supplementary materials the species list? These will help us to understand the species composition of these forests.

We have now added the tree species list in the appendix. We also added the abundance in edge and interior as we think this information can be useful.

Lines 177 – 183: You had 9 transects (in the map caption) how you dived them to get 11 edge sub-transects and 10 interior sub-transects? How long are these transects and sub-transects? Are they at the same length?

We have now added the information in the new section “survey design”

Lines 183 – 194: Did you check the models for multicollinearity? You can use Variance Inflation Factor (VIF)

We did not test for multicollinearity as there is only one factor in the model (edge vs interior)

Lines 200 – 207: Can you add here a table with your distance models and their AICc score? To present your models in order to show also the differences between the distance models.

We have now added this table. Again, we would prefer to avoid overloading the results, so we added them as appendix.

Lines 210 – 212: Can you give the test which was used to find out if there was a significant variation or no? Also, the results of the test?

The test results are in table 2. That is a GAM model.

Line 215: you can not cite a table in the Figure caption. You cand add this explanation in the Results text. You have 33 or 32 vegetation plots?

Thanks for spotting the mistake

Line 219: You have 33 or 32 vegetation plots?

Thanks for spotting the mistake

Lines 222 – 226 can you give also the values of the comparation test? Also, p values, especially for those significantly higher?

See comments above related to stats

Lines 241 – 244: What is the difference between Density of groups and Density of individuals?

Density of groups is just related to the clusters of individuals encountered, density of individuals is considering the group size. This information is present in section 2.4.

Results section: I do not see the GLM results! Have you used this test, as you mention in Methods section?

In table 3 we presented the estimated marginal means based on the GLM, plus we highlighted the p-value when significant. We have now added the estimates, se, F and p-value to the appendix (not in the main text of the ms as again these additional details would break the flow of our narrative).

Lines 251 – 253: Do not repeat the information, just clarify these percent according to my previous comments!

Sentence edited. Thanks.

Lines 332 – 378: You should discuss each lemur species. Compare their density in the study site with other areas and studies and explain why there are differences.

Thanks for your comment. We made an attempt to do that, although we did not want to have a species-based discussion, but rather trying to extrapolate common themes. We think most of the researchers reading this paper would be interested in the broader discussion rather than the species discussion.

Discussion: You can add also some information about species ecology, their preference for different habitats, food, hibernation and others which are suitable to be presented according to your study.

Again we added some of this in the discussion but we would prefer not to load the discussion with species-specific information but rather focus on the broader implication of our findings.

Round 2

Reviewer 2 Report

Dear authors,

     The manuscript Depth of edge influence in a Madagascar lowland rainforest and 2 its effects on lemurs’ abundance was improved, but there are some aspects unclear:

The first one is regarding the Survey design: The difference between each sub transects ai so high? Especially for the interior ones? If the transects do not have the same length, how you can compare them? As you say here, some of your sub-transects are much larger! You have to cut them at the same length and to avoid those observations outside the selected sub-transects. Is normal to have more individuals on a transect of 525 m, compared to one of 925 m.

 

The second one is regarding the GLM. You have only one predictor, which is factorial. You can not analyze such data in GLM! You can use correlation tests, but not GLM! Use other teste for these analyses.

 

Best regards,

Author Response

Dear authors,

     The manuscript Depth of edge influence in a Madagascar lowland rainforest and 2 its effects on lemurs’ abundance was improved, but there are some aspects unclear:

We thank the reviewer for acknowledging that the paper is improved. We hope that with the explaination below, the reviewer will be happy with our methods. 

The first one is regarding the Survey design: The difference between each sub transects ai so high? Especially for the interior ones? If the transects do not have the same length, how you can compare them? As you say here, some of your sub-transects are much larger! You have to cut them at the same length and to avoid those observations outside the selected sub-transects. Is normal to have more individuals on a transect of 525 m, compared to one of 925 m.

We accounted for the different distances of sub-transects in the analysis. We used encounter rate (so number of individuals/km and not count of individuals) as response variable. In addition, we used the length of the transect as weight in the analysis. We have now added a sentence in the data analysis to clarify this.

The second one is regarding the GLM. You have only one predictor, which is factorial. You can not analyze such data in GLM! You can use correlation tests, but not GLM! Use other teste for these analyses.

We are afraid we have to disagree with the reviewer on this point. Correlations (Pearson and Spearman) are usually between scalar variables, not factorial variables. There are options for correlations for ordinal variables and factorials variables as well, but in our case we have a response variable that is scalar and a predictor that is factorial. If one wants to see how a scalar response variable is influenced by a factorial predictor (with two values) the options are Mann-Whitney U test, T-test for independent samples or generalised linear models. Mann-Whitney U tests are suitable for non-normally distributed data but are non-parametric so the power is much lower than parametric tests. T-test cannot be used in our case since the response variable is not normally distributed. Generalised linear models are used to fit the response variable to distributions that are not Gaussian (normal). In our case, we used Tweedie since it was the best fit for our response variables. Mainly, Generalised linear models allow to fit any family to our response variables. We hope to have argued that a Generalised linear model was actually the best options in our context.