Comparing Traditional Methods and Modern Statistical Techniques for Tree Height Prediction
Round 1
Reviewer 1 Report
Comments and Suggestions for Authors
“Combining traditional methods and modern statistical techniques for tree height prediction”
General remarks
1. The title is a bit misleading because the main focus is not on combining but rather on comparing different methods. You might want to consider a slightly altered title.
2. Clearer differentiation between sample trees for calibration and for evaluation (Salzburg, Tyrol) in the beginning of chapter 3 – the evaluation data set could be included as a second data set into the diagram in fig. 2
3. Title/Caption of fig. 2 (“Norway spruce”) should be omitted
4. In section 2.4.2, lines 170-172, you do not express clearly just how many sampling methods you applied: three (random s., stratified s., and cluster s.) or only the latter two?
5. In section 2.4.3, lines 192-193, “all 3 types of data sets”: Do you refer to the (possibly three, see previous comment) data sets resulting from the sampling method? If so, please be more explicit / clearer in your wording.
6. Eq. (2) should be referred to as Michailoff (1943): Zahlenmäßiges Verfahren für die Ausführung der Bestandeshöhenkurven. Forstw. Cbl. 65 (6): 273–279 (primary source).
7. In line 210 you introduce/explain EBLUP as an evaluation parameter. Later (e.g., line 334) you refer to this parameter as EBLUPS. Please harmonize/be more precise.
8. Table 2 / eq.7 (line 314): If coefficient b7 is 0.000, why did you include it into the equation? Maybe you should indicate explicitly that eq.7 gives the general formula and the coefficients are specifically for Picea abies.
9. The results tables 3 and 4 contain absolute values for bias and RMSE. To assess the relevance of these indicators it’d be useful to additionally see the relative values as related to “real” mean y (“percentage RMSE” and “percentage bias”, see for example Pretzsch, J. et al. (2002): The single tree-based stand simulator SILVA: construction, application and evaluation, Forest Ecology and Management 162 (1): 3-21, doi.org/10.1016/S0378-1127(02)00047-6)
10. Diagrams in figs. 3, 4, 9, 10 need more explicit captions on the x-axes
11. Figures 5-6, 11-12: You might want to consider plotting the measured dbh-h pairs as a light grey background which would improve the visual impression of the “fit” of the different models. Not a necessary add-on, but something you could think about.
12. Fig. 7 and others in which you show the reaction of the chosen models to three different dz values: You might add an explanation why you selected 10, 30, an 50cm for dz. They seem to be suitable representatives of certain social / ontogenetic stages (regeneration, intermediary layer, dominant trees). This is not compulsory but might render your choice more logical than just “freely chosen”.
13. Throughout the manuscript, try to avoid the use of “best” as an indicator for model quality, especially with individual parameters such as RMSE. “Highest”/”lowest”, “closest” or “most accurate” / “most precise” are more adequate because “best” already suggests a certain “moral” standard to which you compare your findings. Examples include lines 86, 240, 463, and 491.
14. Be more precise in your wording in the results section, e.g. in line 352: You show the residuals of models derived from / based on / calibrated with the help of stratified data; avoid "... using stratified data" without further explanation. This applies to later instances of the same or similar wording, e.g. line 381, 425, and 502.
15. In the discussion, you compare the models mostly based on RMSE and bias. This is certainly justified in the “central” dbh and height range where there were the majority of measured data available for calibration. The performance of the models at the “edges” (very low and very high values) should nonetheless be reflected with more attention, as an additional quality criterion. This would lead to a slightly more negative evaluation of the random forest models (appropriately, in my opinion) which should be taken up again in the conclusion (lines 739-740).
16. The diagrams in Appendix A should be completed by more descriptive axis captions, e.g. “predicted height (m)” and “observed height (m)”.
Comments for author File: 
Comments.pdf
Author Response
Reviewer 1, thank you for your thoughtful comments. We have considered all your suggestions to the best of our possibilities.
Please see the attachment for more responses.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThis study used different statistical models to predict the tree heights and showed a sound result. However, the current version of the manuscript appears more like a statistics report than an article. The innovations of the paper were not clearly highlighted, and the problems existing in current similar research were not well explained. Therefore, I suggest rejecting the manuscript.
Specific issues:
1. Line 44 - 52. What are the advantages and disadvantages of UHC, GAM & MM methods? Suggest sorting out the current controversies.
2. Line 263 - 273. Just RMSE and Bias can analyze the statistical performance of different methods clearly?
3. Line 314. Table 2. How to filter out the impact of the uncertainty of different methods that use the same volume calculation parameters on the performance of statistical results?
Author Response
Reviewer 2, thank you for your thoughtful comments. We have considered all your suggestions to the best of our possibilities.
Please see the attachment for more responses.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThe document requires conclusions regarding the objectives set and the methodology used. The discussion is very broad and the results are very well illustrated. Expand the summary slightly
Comments for author File: Comments.pdf
Author Response
Reviewer 3, thank you for your thoughtful comments. We have considered all your suggestions to the best of our possibilities.
Please see the attachment for more responses.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsThe authors revise the manuscript according to my comments. It can be accepted now.
