Modeling the Dominant Height of Larix principis-rupprechtii in Northern China—A Study for Guandi Mountain, Shanxi Province
, Lei Zhang 1,2 and Huoyan Zhou 5
John P. Brown
Round 1
Reviewer 1 Report
L20 Change “accurately evaluation” to “an accurate evaluation”
L26-27 This statement is more of a general knowledge situation that can probably be found in most forestry textbooks. Surely those three references weren’t the first to propose these ideas.
L29 Site index is usually stated without the “The” preceding it. Strike.
L35-42 Much of this discussion and even the preceding paragraph can be found in older texts, for instance Clutter, Jerome L., et al. Timber management: A quantitative approach. John Wiley & Sons, Inc., 1983.
L45-L47 Using the word different three times makes this sentence extremely ambiguous as to the meaning it is intended to convey. It seems as though the authors want to suggest there is a hierarchical data structure and data correlations but the discussion hasn’t been presented sufficiently to say that in L47-48. What about the data exactly makes it correlated and hierarchical? That would be the justification for using NLME models.
L50-52. A paragraph needs more than one sentence.
L57 It is stated that there is a lack of relevant height models for larch. In the previous sentence (L55) it is stated that there is a dominant height model for Larix olgensis, which is considered a larch. Maybe don’t use the common name larch here and stick to the scientific names?
L100 The “disc was intercepted’ is unusual wording. Try “Discs were cut at 2 meter intervals above dbh.
L101-L102 I believe the authors mean they cut the stem not the branches (substitute stem for branch in the sentence). Also, I believe they sampled the remaining portion of the stem after 2cm (not m).
L102 Is “analytical wood number” the disc number from the ground up? The phrase is overly scientific.
L104 Use “cut” instead of “intercepted” or even just leave it off here as we know they were already cut. Change “disc” to “discs”,
L107 It would be helpful to many researchers reading this to note to what degree the discs were sanded, meaning, what was the finest sandpaper used, what grit?
L110 Too many “analyticals” in this sentence, particularly the part that says, “…analyze the analytical data…”
L111 What exactly is being corrected with the Carmean method? Are you correcting for sampling in between nodes? If so I am not sure calling that the “tree height deviation” is correct.
L113 This statement regarding suppression is a major problem. Explain what was done here. Is this species of larch shade tolerant and capable of surviving for periods of suppression and then be released? All that is said is that the analytical wood was removed. Have the authors skipped over those rings and did not count them? Site index trees are selected from the dominant trees in the stand because it is hoped they are reflective of a tree that has occupied a dominant position in the canopy since stand initiation. Most of the time shade intolerant species are chosen to measure site index to avoid issues of suppression. See Seymour, Robert S., and Mary Ann Fajvan. "Influence of prior growth suppression and soil on red spruce site index." Northern Journal of Applied Forestry 18.2 (2001): 55-62.
L116 Why aren’t there any height-age pairs below 1.39m if discs were taken from 0, 0.3, and 0.8m heights?
Table 1. It is not clear why there are differing numbers of dominant trees sampled per plot.
Table 2. It would be more appropriate to include the references the models first appeared instead of reference [29].
L148-151 Possibly a symbol is not showing but these equations are written DH having a subscript of i such that the “i” is slightly raised for the second DH term. The value would be zero. L152 attempts to define these terms as the observed dominant height and the estimated dominant height but the variable names are exactly the same. Estimated values typically are defined with a “^” over the term
Were multiple height-age pairs recorded for each disc?
L180-L182, Equation numbering should line up vertically.
L202 “strong fit statistics’ is an imprecise description, reword.
Why is Table 3 separate from Table 2? Notes below a table are supposed to refer to the table above the note(s) and shouldn’t introduce another table.
L204 The statement in parenthesis is not true. Model 9 is very different for RMSE and TRE.
L205 When discussing age class, what were the age classes, 1 year, 10 years?
Figure 3. The legend title should not be “class”. Residuals would be individual points, are these mean standardized residuals for a specific age class?
Table 6: Should use Model 7 and Model 8 not NLME7 and NLME8. “Loglik” is not an appropriate header title.
L235-L236 It’s unclear what the “15 parameter combinations” are. There are 9 models each having 3-4 parameters in the uncorrelated models.
L240 The correlations being observed aren’t really time-series, they are repeated measures on the same subject (tree) as well as plot.
L245 The last term (Zeta) in Eq(18) is not defined. There also shouldn’t be a “1” on mu sub i.
L249 Change “Eq.(19)” to “Eq. (18)”
L258-262 The residuals in Figure 7 are heteroscedastic
L299 “per-showed” is not a word.
L302 “…Model 7, the residual change was too small”. What does this mean?
L320-L321. Not true. In a young stand, two dominant trees may be dominant, that is, dominating their neighbors, but later the two trees may grow large enough to interact. Then one of the two may loose dominance. It was also stated in the methods that some of the trees showed suppression.
Also,
L318-L320. In my opinion, it’s not possible that there were no stumps that couldn’t be attributed to human intervention or natural disaster. Stands self-thin through competition. These stands were from 57-75 years old. There would have been self-thinning occurring as soon as the canopy closed.
Author Response
Comments and Suggestions for Authors
L20 Change “accurately evaluation” to “an accurate evaluation”
Response: thanks for your advice. We modified this in the manuscript. See line 23
L26-27 This statement is more of a general knowledge situation that can probably be found in most forestry textbooks. Surely those three references weren’t the first to propose these ideas.
Response: thanks for your advice. We modified this in the manuscript.
See line 35-37.
L29 Site index is usually stated without the “The” preceding it. Strike.
Response: thanks for your advice. We modified this in the manuscript.
See line 35.
L35-42 Much of this discussion and even the preceding paragraph can be found in older texts, for instance Clutter, Jerome L., et al. Timber management: A quantitative approach. John Wiley & Sons, Inc., 1983.
Response: thanks for your advice. We modified this in the manuscript.
See line 54-65.
L45-L47 Using the word different three times makes this sentence extremely ambiguous as to the meaning it is intended to convey. It seems as though the authors want to suggest there is a hierarchical data structure and data correlations but the discussion hasn’t been presented sufficiently to say that in L47-48. What about the data exactly makes it correlated and hierarchical? That would be the justification for using NLME models.
Response: thanks for your advice. We modified this in the manuscript.
See line 65-74.
L50-52. A paragraph needs more than one sentence.
Response: thanks for your advice. We modified this in the manuscript.
See line 71-73.
L57 It is stated that there is a lack of relevant height models for larch. In the previous sentence (L55) it is stated that there is a dominant height model for Larix olgensis, which is considered a larch. Maybe don’t use the common name larch here and stick to the scientific names?
Response: thanks for your advice. We modified this in the manuscript.
See line 46-47.
L100 The “disc was intercepted’ is unusual wording. Try “Discs were cut at 2 meter intervals above dbh.
Response: thanks for your advice. We modified this in the manuscript.
See line 117.
L101-L102 I believe the authors mean they cut the stem not the branches (substitute stem for branch in the sentence). Also, I believe they sampled the remaining portion of the stem after 2cm (not m).
Response: thanks for your advice. We modified this in the manuscript.
See line 118-119.
Above DBH, we take 2m as the dividing section for sampling. When the shoot of the tree is less than two meters, all of them are taken back to the laboratory for analysis
L102 Is “analytical wood number” the disc number from the ground up? The phrase is overly scientific.
Response: thanks for your advice. We modified this in the manuscript.
See line 119-120.
L104 Use “cut” instead of “intercepted” or even just leave it off here as we know they were already cut. Change “disc” to “discs”,
Response: thanks for your advice. We modified this in the manuscript.
See line 122-123.
L107 It would be helpful to many researchers reading this to note to what degree the discs were sanded, meaning, what was the finest sandpaper used, what grit?
Response: thanks for your advice. We modified this in the manuscript.
See line 125-126.
L110 Too many “analyticals” in this sentence, particularly the part that says, “…analyze the analytical data…”
Response: thanks for your advice. We modified this in the manuscript.
See line 126-128.
L111 What exactly is being corrected with the Carmean method? Are you correcting for sampling in between nodes? If so I am not sure calling that the “tree height deviation” is correct.
Response: thanks for your advice. We modified this in the manuscript.
See line 129.
Height-age curves for individual trees on each plot were very consistent and were combined into an average height-growth curve for each plot. these average plot curves were calculated by averaging tree age at each sectioning height. corrections were also made for a small bias that occurs when height at sectioning point underestimates the actual height attained for that particular year. the bias occurs because sectioning point will always be at some intermediate point along the annual leader rather than at the terminal bud itself. the procedure for removing this small bias is based on the assumption that sectioning point will， on the average， fall in the middle of the annual leaders. Thus, the bias can be removed by increasing height at that section point by one-half the estimated length of the annual leader:
L113 This statement regarding suppression is a major problem. Explain what was done here. Is this species of larch shade tolerant and capable of surviving for periods of suppression and then be released? All that is said is that the analytical wood was removed. Have the authors skipped over those rings and did not count them? Site index trees are selected from the dominant trees in the stand because it is hoped they are reflective of a tree that has occupied a dominant position in the canopy since stand initiation. Most of the time shade intolerant species are chosen to measure site index to avoid issues of suppression. See Seymour, Robert S., and Mary Ann Fajvan. "Influence of prior growth suppression and soil on red spruce site index." Northern Journal of Applied Forestry 18.2 (2001): 55-62.
Response: thanks for your advice. We modified this in the manuscript.
See line 130-134.
L116 Why aren’t there any height-age pairs below 1.39m if discs were taken from 0, 0.3, and 0.8m heights?
Response: thanks for your advice. We modified this in the manuscript.
See line 135.
As the small larch grows rapidly after being unearthed, the growth rate of larch is relatively high within one year after being unearthed, so there is no 0-1.39m data.
Table 1. It is not clear why there are differing numbers of dominant trees sampled per plot.
Response: thanks for your advice. We modified this in the manuscript.
Because our advantage height is defined as 3-4 largest trees in a 400 m2 TSP. Therefore, the sampling quantity of dominant trees in each plot may be different.
Table 2. It would be more appropriate to include the references the models first appeared instead of reference [29].
Response: thanks for your advice. We modified this in the manuscript.
See line 149-153.
L148-151 Possibly a symbol is not showing but these equations are written DH having a subscript of i such that the “i” is slightly raised for the second DH term. The value would be zero.
Response: thanks for your advice. We modified this in the manuscript.
See line 165-173.
L152 attempts to define these terms as the observed dominant height and the estimated dominant height but the variable names are exactly the same. Estimated values typically are defined with a “^” over the term
Response: thanks for your advice. We modified this in the manuscript.
See line 170-174.
Were multiple height-age pairs recorded for each disc?
Response: thanks for your advice. We modified this in the manuscript.
The disc at 0 m is determined according to the age of the tree, and the other discs are determined according to the time required for the tree to grow to the height of the disc.
L180-L182, Equation numbering should line up vertically.
Response: thanks for your advice. We modified this in the manuscript.
See line 197-199.
L202 “strong fit statistics’ is an imprecise description, reword.
Response: thanks for your advice. We modified this in the manuscript.
See line 120-221
Why is Table 3 separate from Table 2? Notes below a table are supposed to refer to the table above the note(s) and shouldn’t introduce another table.
Response: thanks for your advice.
Since the equations shown in Table 3 have initial solutions and are GADA equations, they do not belong to the same category as those in Table 2, so I will separate them.
L204 The statement in parenthesis is not true. Model 9 is very different for RMSE and TRE.
Response: thanks for your advice. We modified this in the manuscript.
See line 221-223.
L205 When discussing age class, what were the age classes, 1 year, 10 years?
Response: thanks for your advice. We modified this in the manuscript.
Since there will be multiple values for the same age, we will average the advantages of the same age, and then calculate the average residual of the same age. Then connect the average values of different ages into a line.
Figure 3. The legend title should not be “class”. Residuals would be individual points, are these mean standardized residuals for a specific age class?
Response: thanks for your advice. We modified this in the manuscript.
Since there will be multiple values for the same age, we will average the advantages of the same age, and then calculate the average residual of the same age. Then connect the average values of different ages into a line.
Table 6: Should use Model 7 and Model 8 not NLME7 and NLME8. “Loglik” is not an appropriate header title.
Response: thanks for your advice. We modified this in the manuscript.
See line 247-248.
L235-L236 It’s unclear what the “15 parameter combinations” are. There are 9 models each having 3-4 parameters in the uncorrelated models.
Response: thanks for your advice. We modified this in the manuscript.
The combination of 15 parameters means that after the introduction of random effects into Model 7 (4 parameters), there are 15 kinds of combinations of random effects in the parameters.
L240 The correlations being observed aren’t really time-series, they are repeated measures on the same subject (tree) as well as plot.
Response: thanks for your advice. We modified this in the manuscript.
In my opinion, the data not only include repeated measurements of the same theme (tree) and graph, but also include tree height (time correlation) of different ages of the same tree. Therefore, this study not only uses the nonlinear mixed effect model (to solve the repeated measurement of the same topic (tree) and graph), but also uses the function (AR) to eliminate the correlation of time series.
L245 The last term (Zeta) in Eq(18) is not defined. There also shouldn’t be a “1” on mu sub i.
Response: thanks for your advice. We modified this in the manuscript.
See line 261.
L249 Change “Eq.(19)” to “Eq. (18)”
Response: thanks for your advice. We modified this in the manuscript.
See line 267.
L258-262 The residuals in Figure 7 are heteroscedastic.
Response: thanks for your advice. We modified this in the manuscript.
See line 280-281.
Since we use SA data, the data is inherently relevant, and we have adopted a variety of functions to eliminate it
L299 “per-showed” is not a word.
Response: thanks for your advice. We modified this in the manuscript.
See line 316-317.
L302 “…Model 7, the residual change was too small”. What does this mean?
Response: thanks for your advice. We modified this in the manuscript.
See line 319-320.
L320-L321. Not true. In a young stand, two dominant trees may be dominant, that is, dominating their neighbors, but later the two trees may grow large enough to interact. Then one of the two may loose dominance. It was also stated in the methods that some of the trees showed suppression.
Response: thanks for your advice. We modified this in the manuscript.
See line 337-345.
Also,
L318-L320. In my opinion, it’s not possible that there were no stumps that couldn’t be attributed to human intervention or natural disaster. Stands self-thin through competition. These stands were from 57-75 years old. There would have been self-thinning occurring as soon as the canopy closed.
Response: thanks for your advice. We modified this in the manuscript.
See line 337-345.
Author Response File: 
Author Response.docx
Reviewer 2 Report
In this manuscript, the site index (SI) models were developed for Larix gmelinii var. principis-rupprechtii using stem analysis data in northern China, Guandishan National Forest Park. In specific, the data of 5,122 height–age pairs from 75 dominant trees in 29 temporary sample plots (TSPs) were collected, and nine conventional SI models were fitted and compared. The random effects were introduced to enhance the conventional models. Overall, this manuscript looks good to me.
However, revisions are still needed to improve the manuscript quality before it could be considered for publication. My major concern is that the introduction is not in a good shape. Specifically, the introduction section is in unorganized logic with many confusing sentences. The innovation of this study is not well highlighted in the introduction section, hence it looks like a “small” case study without sufficient methodology innovation. It missed some background and introduction information, therefore it doesn’t play well as an “introduction” to the general readers. In addition, English writing needs more effort to be improved.
The minor comments are listed below:
1. All the tree samples are from the same national forest park, Guandishan, I am wondering whether they can well represent the “Northern China” as you mentioned in the title. Based on your description in lines 64-66, it seems not. Therefore, suggest adding a subtitle to clarify that this is just a case study.
2. Lines 35-42, all of TSP, PSP, and SA are mentioned as three ways to develop the models. However, you just compared the PSP and SA a little bit, while missed TSP. In addition, just the advantages of SA are described here. Because it doesn’t exist a perfect method without any disadvantages. I highly suggest adding more sentences regarding the disadvantages of SA here, then talk about how you avoid these disadvantages in the method section. This would make your manuscript more robust.
3. Lines 43-52, these two paragraphs need many further efforts. In the current version, the literature review regarding modeling equations is too simple. And the necessity of moving forward the existing models to the proposed “mixed effect model” is not clear and convincing enough. It would be better not to assume the reader know what you know.
Furthermore, there might be some existing studies of growth models using “the mixed effect”, as it’s not something very new. Please double-check and add them if there are.
Therefore, it would be much better to further summarize the literature to have a very clear, organized, and straightforward method-introduction paragraph here.
please combine both paragraphs together if possible.
4. Lines 57-58, you stated that “there is a lack of relevant dominant height models”. Please double-check this sentence to make sure this “absolute” statement is correct. At least, a region, such as Shanxi, Guangdishan, etc. should be added to make this statement more accurate. As I know, there have been some research relevant in the region. For example, “Lars Sprengel, Heinrich Spiecker, Wu Shuirong. Two subject specific modelling approaches to construct base-age invariant polymorphic site index curves with varying asymptotes. Silva Fennica 2022, 56(1): DOI: 10.14214/sf.10544”; and “Sandra-Maria Hipler, Heinrich Spiecker, Wu Shuirong 2021. Dynamic Top Height Growth Models for Eight Native Tree Species in a Cool-Temperate Region in Northeast China. Forests 2021(12): 965 DOI: 10.3390/f12080965”. Just for your information.
5. Lines 84-87, if the plots were randomly selected, how can you make sure the representativeness as you mentioned afterward? According to Table 1, stratified sampling seems used. Please double-check it.
6. Line 125, the ages in table 1 start from 55 rather than 1as said in line 117. So how do you determine the stand age Vs tree age? Used the tree ring to determine the tree age, while used the planting record to know the stand age? Please double-check the sentence.
7. Line 120, suggests adding a national map to show what’s the relative position of these plots to China. Because the international readers maybe don’t know where Shanxi is.
8. In section 2, it would be better to describe what software/modeling platform (with version number) you used to develop the model. R or others?
9. Lines 198-216, did you observe that some models are good in general statistics (Table 5), while they are bad in a certain age range? How do you deal with it if you meet this issue? Just for your consideration.
10. You used both the statistics and the test dataset to validate the models, which is something interesting. The test dataset is usually for Machine Learning methods. The lucky thing is the statistics-based validation and the test-dataset-based validation have relatively consistent conclusions here. If they are not consistent, how do you deal with that? Just for your consideration.
11. Lines 346-353, please re-write the conclusion section to highlight and summarize your key findings of this study here.
12. The English writing still needs more effort to be improved, such as:
in line 9: “accurate” seems more appropriate than “precise”;
in line 28: “easier to be measured” Vs “easier to measure”;
in line 61: “norther” Vs. “northern”;
in line 73: “cover level” Vs. “cover rate”;
…
Please carefully go through the manuscript to avoid grammar issues.
Author Response
Comments and Suggestions for Authors
In this manuscript, the site index (SI) models were developed for Larix gmelinii var. principis-rupprechtii using stem analysis data in northern China, Guandishan National Forest Park. In specific, the data of 5,122 height–age pairs from 75 dominant trees in 29 temporary sample plots (TSPs) were collected, and nine conventional SI models were fitted and compared. The random effects were introduced to enhance the conventional models. Overall, this manuscript looks good to me.
However, revisions are still needed to improve the manuscript quality before it could be considered for publication. My major concern is that the introduction is not in a good shape. Specifically, the introduction section is in unorganized logic with many confusing sentences. The innovation of this study is not well highlighted in the introduction section, hence it looks like a “small” case study without sufficient methodology innovation. It missed some background and introduction information, therefore it doesn’t play well as an “introduction” to the general readers. In addition, English writing needs more effort to be improved.
Response: thanks for your advice. We modified this in the manuscript.
See line 28- 84.
We have revised the grammar of the whole manuscript.
The minor comments are listed below:
- All the tree samples are from the same national forest park, Guandishan, I am wondering whether they can well represent the “Northern China” as you mentioned in the title. Based on your description in lines 64-66, it seems not. Therefore, suggest adding a subtitle to clarify that this is just a case study.
Response: thanks for your advice. We modified this in the manuscript.
See line 28- 84.
The research in this study can only represent the Guandi Mountain area. Due to my own writing problems, such errors have been corrected.
- Lines 35-42, all of TSP, PSP, and SA are mentioned as three ways to develop the models. However, you just compared the PSP and SA a little bit, while missed TSP. In addition, just the advantages of SA are described here. Because it doesn’t exist a perfect method without any disadvantages. I highly suggest adding more sentences regarding the disadvantages of SA here, then talk about how you avoid these disadvantages in the method section. This would make your manuscript more robust.
Response: thanks for your advice. We modified this in the manuscript.
See line 54-66, 192 -209.
- Lines 43-52, these two paragraphs need many further efforts. In the current version, the literature review regarding modeling equations is too simple. And the necessity of moving forward the existing models to the proposed “mixed effect model” is not clear and convincing enough. It would be better not to assume the reader know what you know.
Response: thanks for your advice. We modified this in the manuscript.
See line 42-75.
Furthermore, there might be some existing studies of growth models using “the mixed effect”, as it’s not something very new. Please double-check and add them if there are.
Response: thanks for your advice. We modified this in the manuscript.
See line 70-76.
Therefore, it would be much better to further summarize the literature to have a very clear, organized, and straightforward method-introduction paragraph here. please combine both paragraphs together if possible.
Response: thanks for your advice. We modified this in the manuscript.
See line 28-84.
- Lines 57-58, you stated that “there is a lack of relevant dominant height models”. Please double-check this sentence to make sure this “absolute” statement is correct. At least, a region, such as Shanxi, Guangdishan, etc. should be added to make this statement more accurate. As I know, there have been some research relevant in the region. For example, “Lars Sprengel, Heinrich Spiecker, Wu Shuirong. Two subject specific modelling approaches to construct base-age invariant polymorphic site index curves with varying asymptotes. Silva Fennica 2022, 56(1): DOI: 10.14214/sf.10544”; and “Sandra-Maria Hipler, Heinrich Spiecker, Wu Shuirong 2021. Dynamic Top Height Growth Models for Eight Native Tree Species in a Cool-Temperate Region in Northeast China. Forests 2021(12): 965 DOI: 10.3390/f12080965”. Just for your information.
Response: thanks for your advice. We modified this in the manuscript.
See line 45-49.
- Lines 84-87, if the plots were randomly selected, how can you make sure the representativeness as you mentioned afterward? According to Table 1, stratified sampling seems used. Please double-check it.
Response: thanks for your advice. We modified this in the manuscript.
See line 99-106.
- Line 125, the ages in table 1 start from 55 rather than 1as said in line 117. So how do you determine the stand age Vs tree age? Used the tree ring to determine the tree age, while used the planting record to know the stand age? Please double-check the sentence.
Response: thanks for your advice. We modified this in the manuscript.
First, we conduct a survey on each tree of larch in the forest, and then select the average tree in the forest, and use the growth awl to get the age. The average age of several average trees is the age of the forest. The data taken is a disk, and the growth data (1-75years) of the tree is obtained through Xlstem software.
Table 1 shows the average age of stands.
- Line 120, suggests adding a national map to show what’s the relative position of these plots to China. Because the international readers maybe don’t know where Shanxi is.
Response: thanks for your advice. We modified this in the manuscript.
See line 138-139.
- In section 2, it would be better to describe what software/modeling platform (with version number) you used to develop the model. R or others?
Response: thanks for your advice. We modified this in the manuscript.
See line 213-214.
- Lines 198-216, did you observe that some models are good in general statistics (Table 5), while they are bad in a certain age range? How do you deal with it if you meet this issue? Just for your consideration.
Response: thanks for your advice.
If the fitting effect in a certain age group is bad, first check the data and delete the outliers in the data. If the fitting effect is improved, accept it. If the fitting effect is still poor, use the second step. Step 2: For this age segment, select appropriate stands for re sampling, and then analyze and fit the model.
You used both the statistics and the test dataset to validate the models, which is something interesting. The test dataset is usually for Machine Learning methods. The lucky thing is the statistics-based validation and the test-dataset-based validation have relatively consistent conclusions here. If they are not consistent, how do you deal with that? Just for your consideration.
Response: thanks for your advice.
There are many possible reasons for inconsistency, which may be caused by the large variability of data. It is recommended to replace the model or check the data. Or use another test method to leave a cross validation to test the variability of the data.
- Lines 346-353, please re-write the conclusion section to highlight and summarize your key findings of this study here.
Response: thanks for your advice. We modified this in the manuscript.
See line 368-380.
- The English writing still needs more effort to be improved, such as:
in line 9: “accurate” seems more appropriate than “precise”;
in line 28: “easier to be measured” Vs “easier to measure”;
in line 61: “norther” Vs. “northern”;
in line 73: “cover level” Vs. “cover rate”;
Please carefully go through the manuscript to avoid grammar issues.
Response: thanks for your advice. We modified this in the manuscript.
We have revised the grammar of the whole manuscript.
Author Response File: Author Response.docx
