Forest Biometric Systems in Mexico: A Systematic Review of Available Models
, Benedicto Vargas-Larreta 3, Edgar J. González 4,*, José Javier Corral-Rivas 5, Oscar A. Aguirre-Calderón 6, Eduardo J. Treviño-Garza 6, Héctor M. De los Santos-Posadas 7, Martin MartÃnez-Salvador 8, Francisco J. Zamudio-Sánchez 9 and Cristóbal Gerardo Aguirre-Calderón 3
Round 1
Reviewer 1 Report
In the current article, a review of the available forest models for various variables at tree level in Mexico is attempted. The availability and the implementation of similar models are essential in the context of Sustainable Forest Management (SFM), hence I find the article really interesting. However, there are some issues that need to be addressed. Those are mentioned below.
- L29-49. Please short the abstract to 300 words maximum, following the journals’ requirements.
- L42. Please improve the expression.
- L46. I suggest authors avoid the first person throughout the manuscript.
- L57. Perhaps “dendrometric” instead of “dasometric” is better suited (variables at tree level).
- L77 – 76. R-squared provides the advantage of a measure of the total variance (of the dependent variable) that is currently explained by the model. However, sometimes it is invalid for nonlinear models and many authors have renamed it as “efficiency index”, among other names. Furthermore, in the current manuscript, it has been used as the main measure of a good fit. This is clearly a shortage and has to be mentioned (briefly).
- L85-86. Please improve the expression.
- L112. This must be transferred to the relevant section (L347).
- L159. Please improve the figure’s quality (it is very informative though).
- L166. Please replace “BHD” with “DBH”.
- L176. Please set “3” as an exponent.
- L189. Please improve the figure’s quality.
- L207. Please see the previous comment.
- L287. Please see the previous comment.
- L304-305. Please see the fifth comment. In addition, the model's biological explanation should be added as one basic criterion of good fit.
- L315. In L242 the MSE was mentioned as the main error measure. Which is the correct one?
- L317. Please remove RMSE from this line.
- L319-322. Those lines are incomplete. Firstly, I suggest authors include BIC as an additional measure. Similar to AIC, it is used for nested models comparison. The (approximate for nonlinear forms) t-statistic of the parameter estimates is an additional criterion that finally determines the included variables and the model’s complexity. They must be mentioned both.
- L328-336. Somehow, the issues raised in the discussion section (validation) are not mentioned in the results. Please, improve the specific section.
Author Response
General comments
In the current article, a review of the available forest models for various variables at tree level in Mexico is attempted. The availability and the implementation of similar models are essential in the context of Sustainable Forest Management (SFM), hence I find the article really interesting. However, there are some issues that need to be addressed. Those are mentioned below.
Answer: Thank you very much for your kind comments, they are very gratifying for us.
Specific comments
- L29-49. Please short the abstract to 300 words maximum, following the journals’ requirements.
Answer: Done, abstract was adjusted to 300 words maximum, following the journals’ requirements.
- L42. Please improve the expression.
Answer: We are unsure about the reviewer’s request. Since we modified the abstract, we hope this request has been fulfilled.
- L46. I suggest authors avoid the first person throughout the manuscript.
Answer: We are unsure if this is a necessary modification, since it is a common practice to use first person in regular scientific English (https://twp.duke.edu/sites/twp.duke.edu/files/file-attachments/first-person.original.pdf). Indeed, we sought the latest published articles in Forests and we believe that the use of the first person is admitted since several articles included it (https://www.mdpi.com/1999-4907/13/4/600, https://www.mdpi.com/1999-4907/13/4/590, https://www.mdpi.com/1999-4907/13/4/599). However, we have no problem if the Editor considers that it is important to eliminate the first person of the manuscript.
- L57. Perhaps “dendrometric” instead of “dasometric” is better suited (variables at tree level).
Answer: Done; we changed the word “dasometric” for “dendrometric” in L30 and eliminated it to avoid confusions in L59.
- L77 – 76. R-squared provides the advantage of a measure of the total variance (of the dependent variable) that is currently explained by the model. However, sometimes it is invalid for nonlinear models and many authors have renamed it as “efficiency index”, among other names. Furthermore, in the current manuscript, it has been used as the main measure of a good fit. This is clearly a shortage and has to be mentioned (briefly).
Answer: We agree with the reviewer in that R2 is not the most valid goodness-of-fit measure. However, since the objective of this study was to perform a systematic review of the existing literature on biometric systems, we favored in our analyses the most frequently used metric, i.e., R2. To focus on another measure would have substantially reduced the number of cases analyzed. We included a line at the end of Section 2.3 to recognize this fact (L156-157).
- L85-86. Please improve the expression.
Answer: Done. The sentence “The search for information focused on volume, height, site index and diameter equations that evaluated individual trees or stands, published before 2020” was changed to “The search for information focused on volume, height, site index and diameter equations constructed from individual trees or stands data, and published before 2020.” in L91-92. We hope this change made the expression clearer.
- L112. This must be transferred to the relevant section (L347).
Answer: We beg to differ with the reviewer in transferring this information elsewhere, as Items 8 and 9 of the PRISMA checklist request this information to be provided in the Methods section.
- L159. Please improve the figure’s quality (it is very informative though).
Answer: The original figures were uploaded with high resolution to the Journal’s platform.
- L166. Please replace “BHD” with “DBH”
Answer: Done. The word “BHD” was changed to “DBH”, as well as its meaning (L177).
- L176. Please set “3” as an exponent.
Answer: Done.
- L189. Please improve the figure’s quality.
Answer: The original figures were uploaded with high resolution to the Journal’s platform.
- L207. Please see the previous comment.
Answer: The original figures were uploaded with high resolution to the Journal’s platform.
- L287. Please see the previous comment.
Answer: The original figures were uploaded with high resolution to the Journal’s platform.
- L304-305. Please see the fifth comment. In addition, the model's biological explanation should be added as one basic criterion of good fit.
Answer: We agree with the reviewer and have included “make biological sense” as another criterion of goodness-of-fit in L359-360.
- L315. In L242 the MSE was mentioned as the main error measure. Which is the correct one?
Answer: Indeed this was a mistake from our part; we have changed the term with RMSE, which is the correct term.
- L317. Please remove RMSE from this line.
Answer. Done. We also eliminated R2 for consistency.
- L319-322. Those lines are incomplete. Firstly, I suggest authors include BIC as an additional measure. Similar to AIC, it is used for nested models comparison. The (approximate for nonlinear forms) t-statistic of the parameter estimates is an additional criterion that finally determines the included variables and the model’s complexity. They must be mentioned both.
Answer. We beg to differ with the reviewer in that BIC and the t-statistic are additional measures for model comparison (i.e., model selection). Both these measures aim at hypothesis testing (Aho et al. 2014 Ecology 95:631; De Wayne et al. 2017 The American Statistician 72:379). In constructing biometric systems we aim at predicting forest and tree attributes from sets of variables, not testing hypotheses on how these attributes are biologically determined by a given set of variables. Therefore, we decided to keep AIC as our favored measure for model selection. As we discussed after the introduction of AIC, even if we were to use AIC, such use does not go into the main problem of not evaluating predictive power when performing model selection with the aim of having predictive models.
- L328-336. Somehow, the issues raised in the discussion section (validation) are not mentioned in the results. Please, improve the specific section.
Answer: We recognize this missing aspect in our study. We have included a few sentences concerning validation and prediction both in the Introduction (L74-81), Methods (L130-131) and Results (L314-316) sections
Reviewer 2 Report
Dear Authors,
thank you very much for your manuscript. I like to produce some comments:
a) for the selection of the most performancess model you use "R2". In the section 253-263 appear as you selected the Shumamacher and Hall becouse their model has been used more than the others;
b) 50-year is very long time and during this period many thing was change. Example the tools for measure the forest parameters. Do you know if this can have an effect on the model performances???
c) I am agree about your assumption mentioned at rows 46-47 and 339, but since the globalization and climate change processes, maybe forests have changed and also the increment for year of volume, diameter and heigh are changed. Can do you explain about this point?
Last issues
row 295: I suggest "Discussion and Conclusion
row 311: the use of ")" is not clear
Author Response
Thank you very much for your manuscript. I like to produce some comments:
- For the selection of the most performance model you use "R2". In the section 253-263 appear as you selected the Shumamacher and Hall because their model has been used more than the others.
Answer: We are unsure about the meaning of the reviewer’s comment but we must emphasize that we are not performing model selection based on R2. We only report the frequency with which models are used in the literature.
- 50-year is very long time and during this period many thing was change. Example the tools for measure the forest parameters. Do you know if this can have an effect on the model performances???
Answer: We agree with the reviewer’s observation that tools have changed over the last 50 years. However, we must assume that they had no effect on model performance since this element was not considered in the design of this study.
- I am agree about your assumption mentioned at rows 46-47 and 339, but since the globalization and climate change processes, maybe forests have changed and also the increment for year of volume, diameter and heigh are changed. Can do you explain about this point?
Answer: Globalization and climate change processes have been occurring over the span of decades and their influence on forest attributes must already be reflected in the data of each case of our study. However, this is beyond the scope of the objectives of our study and its evaluation would require access to the original data.
Last issues
row 295: I suggest "Discussion and Conclusion
Answer: Since the inclusion of a conclusion seemed optional for the Journal, we did not include one in our manuscript. However, if the Editor considers it necessary, we could include it.
row 311: the use of ")" is not clear
Answer: We eliminated this typing error.
Round 2
Reviewer 1 Report
The revisions that authors made to the manuscript are very effective in addressing my concerns. My decision is that the manuscript is significantly improved.
