Round 1

Reviewer 1 Report

I read the paper carefully. The authors tried to integrate different kind of variables and models to model and predict the distribution of Swietenia macrophylla. Generally the paper structured well and the text is coherent. But regarding the methodology and results I think it needs some more works before publishing. I also have some concerns and suggestion, which you can find them below:

Line 181: Increase the quality of the Fig 1. The grid isn’t readable. Use geographic coordinate system and improve the resolution.

Line 174. In the section “study area” add more information such as elevation gradient, dominated species, climatology information and etc.

Table-1 (line 225). Why you didn’t consider more variables? Such as topographic variables. Do you think only considering some bioclimatic and soil variables can provide a reliable model?  In short try to use more variables or clarify why you used these variables. 

Line 230. Why you used the default values? Why you didn’t consider the hyperparameter tuning process to determine the best combination of parameters and increase the learning process? Did you use cross validation? Add the relevant information here, also present the optimal values for the hyperparameters of the used models.  

How about the most important variables. What variables have a greater impact on your models? Present the variable importance scores in the results, also the relevant information in the material and methods.

Author Response

Dear reviewers:

On behalf of all the authors of the article entitled: Modelling climatically suitable areas for Mahogany (Swietenia macrophylla King) and their shifts across Neotropics: the role of protected areas; I appreciate your kind comments and suggestions, as they have allowed us to improve the scientific quality of the manuscript. Below we present in detail and by number of lines the changes made to the text:

Reviewer 1

Review 1: I read the paper carefully. The authors tried to integrate different kind of variables and models to model and predict the distribution of Swietenia macrophylla. Generally the paper structured well and the text is coherent. But regarding the methodology and results I think it needs some more works before publishing. I also have some concerns and suggestion, which you can find them below:

Authors: Dear reviewer, we appreciate your valuable comments and suggestions for the improvement of our manuscript.

Review 1: Line 181: Increase the quality of the Fig 1. The grid isn’t readable. Use geographic coordinate system and improve the resolution.

Authors: Dear reviewer. Thank you for your valuable observation. We have improved the quality and resolution of the Figure and changed the coordinate system. The modifications made can be seen in Figure 1 and Line 184.

Review 1: Line 174. In the section “study area” add more information such as elevation gradient, dominated species, climatology information and etc.

Authors: Dear reviewer, we have included the requested information. The changes made can be seen on the following lines 179-183.

Review 1: Table-1 (line 225). Why you didn’t consider more variables? Such as topographic variables. Do you think only considering some bioclimatic and soil variables can provide a reliable model?  In short try to use more variables or clarify why you used these variables.

Authors: Dear reviewer, thank you for your comment. We agree that it could be possible to include more variables in order to obtain a more precise model (closer to the real distribution of the species) for the present. But, the objective of this work is to generate models for present and future climate scenarios. In the case of future predictions, it is not recommended to use topographic variables, such as altitude, orientation, slope, etc. Since they are indirect variables of the climate (temperature and precipitation), it is preferable to use variables directly related to the niche of the species (such as climate or soil) that allow reflecting the niche of the species correctly and therefore making correct temporal extrapolations [1]. We are modelling the fundamental niche of the species, i.e. the potential distribution of the species, we can not model the real distribution of the species.

Review 1: Line 230. Why you used the default values? Why you didn’t consider the hyperparameter tuning process to determine the best combination of parameters and increase the learning process? Did you use cross validation? Add the relevant information here, also present the optimal values for the hyperparameters of the used models. 

Authors: Dear reviewer, we consider that your concern was of great relevance for the improvement of our manuscript so we have implemented a new methodology that considers the evaluation of the modification of the parameters in the modelling algorithms. The new methodology implemented is detailed in the following lines 239-244.

Review 1: How about the most important variables. What variables have a greater impact on your models? Present the variable importance scores in the results, also the relevant information in the material and methods.

Authors: Dear reviewer, thank you very much for your valuable comments. Based on your suggestions we have incorporated contributions in the materials and methods section on lines 251-252. Additionally, in the results section on lines 286-288.

Reviewer 2 Report

See the attachment for detailed comments.

Comments for author File: Comments.pdf

Author Response

Reviewer 2

Reviewer 2: I highly recommend using a method like convex hulls or buffers to select background points, if the entire region in Figure 1 was used as the background for model development. Using too large a background area will allow for inaccessible areas to be included in model development, leading to AUC score inflation and erroneous models.

Authors: Dear reviewer, thank you for your opinion. It could be true in certain situations. However, the pseudoabsences weights were calculated such that prevalence = 0.5, in order to avoid this problem. On the order hand, it has been recommended for some authors the reverse strategy, i.e. not select the pseudo-absences points that are not too close to presences (it would avoid the same niche and pseudo-replication, [1,2]). Furthermore, it is preferable to represent as possible the whole climate available in the study area in other to generate correct temporal extrapolations [3].

Reviewer 2: Why was 2.5arcmin selected and not 30arcsec? The finer resolution will generate better models and allow for more precise predictions. All environmental data used in these analyses are available at 30arcsec.

Authors: Dear reviewer, we share your concern. Initially, we ran the models at a resolution of 2.5arcmin for computational reasons. However, due to your valuable suggestion we have decided again to generate the models considering a resolution of 30arcsec. The modifications made to the text were included in the line 215-217.

Reviewer 2: Using the ensemble model in BioMod2 is excellent. I do recommend testing other sets of model parameters other than just default though. The default parameters are very general and need to be fine-tuned.

Authors: Dear reviewer, we consider that your concern was of great relevance for the improvement of our manuscript so we have implemented a new methodology that considers the evaluation of the modification of the parameters in the modelling algorithms. The new methodology implemented is detailed in the following lines 239-244.

Reviewer 2: I recommend adding in other evaluation metrics (eg, AICc, CBI, Omission Rate) other than just AUC

Authors: Dear reviewer, we appreciate your valuable suggestions. We have incorporated 3 additional evaluation metrics. The modifications made to the text can be seen in the following lines 254-255.

Reviewer 2: Was this resampled to the same resolution as the input variables for the ENM? If not, they should be resampled to the same resolution

Authors: Dear reviewer, thank you for your valuable suggestions. Indeed the variables were resampled to the resolution used in the elaboration of the models. This consideration can be seen in line 274.

Reviewer 2: I think it is important to note that these models predict where suitable habitat (per the input variables) will be, not necessarily where the species will be. Some of the results is framed well in this context, but the title of this section should be rephrased to keep consistent.

Authors: Dear reviewer, thank you very much for your suggestions. We share your criteria and indeed in terms of distributional ecology the models made with predictions of possible environmentally suitable areas where the species could be found. For this reason, we have modified the text in question and the changes made can be seen in line 305

Dear reviewers, thank you for your valuable input to our manuscript; we have modified all your comments. In addition, we have included your valuable contribution in the acknowledgments as a token of gratitude.

References

1. Guisan, A.; Thuiller, W.; Zimmermann, N.E. Habitat Suitability and Distribution Models: With Applications in R; Cambridge University Press, 2017; ISBN 0521765137.
2. Thuiller, W.; Lafourcade, B.; Engler, R.; Araújo, M.B. BIOMOD–a Platform for Ensemble Forecasting of Species Distributions. Ecography (Cop.). 2009, 32, 369–373.
3. Chevalier, M.; Zarzo-Arias, A.; Guélat, J.; Mateo, R.G.; Guisan, A. Accounting for Niche Truncation to Improve Spatial and Temporal Predictions of Species Distributions. Front. Ecol. Evol. 760.

Round 2

Reviewer 1 Report

The paper is now ready for publication.