A New Forest Growing Stock Volume Estimation Model Based on AdaBoost and Random Forest Model
Round 1
Reviewer 1 Report (Previous Reviewer 2)
Comments and Suggestions for AuthorsGeneral comments:
While the authors revisions improved the manuscript, issues still remain.
The authors still confuse precision with accuracy. They are not the same. A simple google search reveals:
Accuracy measures how close results are to the true or known value. Precision, on the other hand, measures how close results are to one another.
The authors absolutely need to conform to traditional meanings of accuracy and precision.
It would be helpful if the authors could expand on how the method could be employed in a practical sense. Would this be used to substitute for future ground surveys? To supplement them during periods between inventories? Or to interpolate between ground survey points? If the answer is yes to all, then which ones would be strongest and which ones weakest?
Specific comments:
38 I do not see how anyone could possibly determine the accuracy of any inventory method unless one actually weighed the amount of biomass or determined the exact volume of each tree in an area. It is literally impossible to do. What one can certainly determine is the precision of an inventory. That is how variable the estimates of the mean are. This is not the same as accuracy.
42 Again, impossible to know how accurate the survey/inventory is. It is relatively easy to determine the precision.
61 I do not see how spectral information itself can have a negative correlation. That is like saying light has a negative correlation to something. Increasing IR might have a negative correlation or decreasing IR might have a negative correlation, but IR cannot on its own have a correlation. For there to be a correlation there must be a trend of some sort in the two variables.
134 Do the authors mean the growing stock volume is estimated to be as high as the value indicated? And what is the source of this information?
156-160 select at the start of each sentence is not needed. It is repetitive and unnecessary.
189 I am not sure the heading methods makes sense here. Are these specific kinds of methods? They seem to be AI-related methods.
276-283 This all seems like methods and should be in the methods section.
350 This cannot be Figure 2. Most likely Figure 4.
366 The units on the Y axis are not defined. They need to be defined somewhere: on the Y axis, in the Figure title, or next to the variable in each panel. This is standard journal practice.
379 As with the earlier figure the units associated with the Y-axis are not defined. They need to be defined somewhere. Also it is not clear from the figure or the text what the index is. Index of what exactly? Without knowing this the reader cannot understand the figure.
460 I am not sure what simpler means in this context. Do the authors mean more limited range of these variables? A more precise statement would be more helpful to the reader.
484 see earlier comments on accuracy versus precision.
501 Precision?
Author Response
Thank you very much for revisiting my manuscript. Your commitment to scientific rigor is truly commendable, and I value your guidance.
I have carefully revised my manuscript point by point according to your comments.
Please see the attachment
Author Response File: 
Author Response.pdf
Reviewer 2 Report (Previous Reviewer 1)
Comments and Suggestions for AuthorsReview of manuscript
A New Forest Growing Stock Volume Estimation Model Based on AdaBoost and Multi-Source Data
by Chao Zhang et al.
General comments:
This work focuses on the RF-Adaboost model, a unique predictor of forest expanding stock volume. This paper examines the performance of deep learning methods using several constructed datasets. An updated draft of the manuscript indicates that the article might appear in the journal Forests
Specific comments:
Line 21: ... optimal performance... Specify what you are modeling.
Line 50: ... feature parameters... Probably stand size parameters or variables.
Line 52: ... is not very high... Base on the values ​​of statistical indices.
Line 254: Explain how the observed values ​​y were calculated and to what degree of precision.
Line 257: N‘ isn‘t defined.
Line 259: The function ht(x) is not defined.
Line 262: Explain the maximization procedure in Equation 3, specifying which variable is being maximized.
Eq. (10): In the text, MAPE is written as a percentage, so the right side of equation 10 must be multiplied by 100.
Lines 520 – 532: References do not meet journal requirements.
The article is written in acceptable language.
Author Response
Thank you very much for revisiting my manuscript. Your commitment to scientific rigor is truly commendable, and I value your guidance.
I have carefully revised my manuscript point by point according to your comments.
Please see the attachment
Author Response File: Author Response.pdf
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsReview of manuscript
A New Forest Growing Stock Volume Estimation Model Based on AdaBoost and Multi-Source Data
by Chao Zhang et al.
General comment:
This paper focuses on examining the performance of deep learning methods using several constructed datasets.
1. The "volume of forest growing stock" is a dynamic process, but the deep learning methods used are static. How to justify the practical usefulness of the models?
2. The models presented here are described in detail in previous publications on deep learning. However, the authors have decided that it is possible to carry out the datasets, adapt the models and prepare the paper.
3. The abstract and introduction do not highlight the novelty of the work and its contribution to the field of growth models in forestry.
4. The models presented in this paper are not comparable to conventional regression growth models and therefore the introduction does not present their advantages and limitations. Why are regression models eliminated when models are fitted to existing data for comparison?
In view of this, I propose that this article be substantially revised for publication in Forests.
Comments on the Quality of English LanguageThe article is written in acceptable language
Author Response
Thank you very much for revisiting my manuscript. Your commitment to scientific rigor is truly commendable, and I value your guidance.
I have carefully revised my manuscript point by point according to your comments.
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsGeneral comments
While the manuscript explores some interesting facets of estimating the growing stock volume, it also contains a number of problems that need to be addressed.
While the authors refer to improved accuracy of their approach I don’t think this what was explored. Accuracy is an indication of how close one is to the true value. Unfortunately there is no way for anyone to assess this because no one knows the true volume, not at this scale. I suppose one could examine the precision of the method or its relative precision, but that is not the same as accuracy.
I understand that one way to assess how a model works is to divide the dataset up using one part for model development and one for testing. However, this method is very limited in telling one how well the model works for a truly novel data set. This limitation needs to be discussed particularly if the proposed method is to predict what an independent dataset contains. In other words, the models described may perform well for the dataset they were developed from. The real question is whether they work well for other datasets. If not, then I don’t see how the method could possibly substitute for actual inventories in the future as seems to be suggested. The authors need to discuss the limitations of the method being used. Perhaps the method might be used to interpolate for a given inventory dataset? Perhaps it might not be particularly useful when underlying relationships have changed or were likely to have changed?
Specific comments (line)
11 field surveys?
13-16 All the methods should be in past tense. That is true for the results as well.
34 These terms seem to apply to China, but other terms are used elsewhere. Either generalize the terms or make it clear where it applies.
72 researchers?
80 textures of what exactly?
96 do the authors mean machine learning models or models of all sorts?
114 One has to get to the discussion to learn the objectives of the study. Shouldn’t that be made clear in the introduction?
130 some changes in font sizes at this point.
133 Unfortunately there are two maps in this figure. While the reader can probably figure which one shows the districts the description should be clearer.
139-140 What is the source of this information? Is this generally known?
152-155 There is no need to repeat “select sample sites” as this is already in the preceding sentence.
199 less than what exactly? It must be compared to something to make sense.
280 while this table is repeatedly referred to for the number of variables, I don’t see that information in the table. Also, I don’t see how these are parameters. Maybe the problem is that there are two Table 3s.
300 What are the Y-axes on the figure?
326 What are the Y-axes on the figure?
345 What are the Y-axes on the figure?
349 random forest?
374 What is the index? Index of what?
378 The study should be described in past tense, not present tense
428 These do not seem to be methods as much as different studies. Did they use different methods? It is not clear as written.
431 What evidence is there that these differences are significant? In which sense?
469-474 This paragraph seems like conclusions. Perhaps these sentences should be moved to this section and combined with like ideas in the conclusions?
Comments on the Quality of English LanguageSee comments to authors
Author Response
Thank you very much for revisiting my manuscript. Your commitment to scientific rigor is truly commendable, and I value your guidance.
I have carefully revised my manuscript point by point according to your comments.
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThe paper presents a model for estimating forest growing stock volume based on multi-source data and machine learning methods. The authors use Landsat 8 satellite data and a combination of Random Forests and Adaboost methods. The problem of forest growing stock volume estimation by remote sensing methods is relevant, but the research of this paper is very similar to the work of H. Huang DOI: 10.3390/f13091471. H. Huang also uses remote sensing data, vegetation indices, four data combination schemes, and three machine learning algorithms, with a combination of radar Sentinel-1 and spectral Sentinel-2 data, which has a higher resolution of 10 m, compared to 30 m of the Landsat-8. Sentinel-2 data are also more suitable for forest surveys because of the larger number of bands, including additional bands in the red edge region of the spectrum, very important for vegetation separation.
Unfortunately, therefore, I do not see any advantages in this study - repeating the work of H. Huang on coarser data with a different machine learning algorithm.
Additionally, the Discussion does not quite correctly compare the results with other works. Firstly, the table 8 (line 428) is taken entirely from the work of H. Huang (DOI: 10.3390/f13091471). Secondly, the study in 2023 should be compared with more recent results, and not with the work of 2018-2019. Also, in the work H. Huang used data with a higher resolution of 10 m, so comparing them with the results of work at a resolution of 30 m is not correct.
In the Introduction there is no justification why the authors chose Adaboost methods to study forests.
Incorrect numbering of tables, table 3 is repeated twice.
