Satellite Assessment of Forest Health in Drought Conditions: A Novel Approach Combining Defoliation and Discolouration
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsIn general, the manuscript is well-written and has a clear storyline. The two-phase sampling approach is well-justified and methodologically appropriate. However, I do have a few minor comments that could enhance the general readability of the manuscript:
1- Including more references to similar studies and their findings could provide better context and justification for the study.
2- The methods and results are dense and technical. Simplifying the language and breaking down complex sentences could improve readability for a broader audience.
3- While the study’s findings are significant, a deeper analysis of the implications for forest management and policy would add value. This information can be incorporated into the discussion section.
Author Response
We thank the reviewer for careful reading of the manuscript and constructive remarks. We have taken all comments into account to improve and clarify the manuscript. Please find in the attachment a response to all comments (our replies are in dark blue colour).
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThis paper deals with the issue of finding forest stress conditions in optical imagery (sentinel 2, Landsat 9) in Slovakia, using field work at a number of study sites as a means of verifying any results. I think two things I would highlight which maybe could also be more prominent in the summaries of the paper - the assessment of 67% of the forest receiving at least moderate damage is a good result (in a scientific sense - it is of course disturbing for biosphere health) that should be more prominent. Likewise I think the discussion of the confounding of discolouration and defoliation (and proposed combined effect measure DEFxDIS) is a useful one and could be more prominent in any summary parts of the text.
The paper is well written, and a good contribution to this topic that should be published. Overall the results are well presented, I have a few comments below, but they are relatively minor. Thanks for writing the paper it was very interesting to read.
Comments:
A general point it would be helpful to clarify is - what sort of forests is this methodology best applied to? It seems largely to be concerned with coping with mixed stands of deciduous/coniferous - but would this work for other types of mixed species or less continuous forest? Is it particularly appropriate for temperate forests - will it work elsewhere etc. It might be helpful to give a little more of a view on this.
L127 - I don't understand the comment about using Landsat 9 to in-paint the cloud regions in the Sentinel 2 image. There is a slight band mismatch (is this significant?) and a resolution mismatch. It would be helpful to briefly mention how these were dealt with.
L140 - The table gives bands and resolutions for Sentinel 2, but these don't exactly match up to any Landsat 9 parts of the mosaic image. It might be better not to give these if they are only valid for part of the image, again useful to clarify.
L245 - It might be worth giving a quick heuristic description for equation 1. I.e defoliated leaves cannot be discoloured, so DIS as a percentage only affects the leaves left on the tree when DEF is accounted for. I think this is explained through the text, but could be a bit more clearly stated here. L248-257 sort of explain this, but they are not very clear.
Figure 4 - These plots are really hard to read in 3d, it isn't clear to see (L296) the NSC1 component insensitivity, I can really only 'kind-of' tell that the plot is flattened in that direction. These might be better replaced or supplemented by equivalent marginal plots, or contour plots.
L320 - Your result plots classify some pixels as non-forest, but the reference to the forest mask you use [53] is a little buried here. It would be helpful to make a little more comment on this mask and maybe to discuss dependence on the mask used - what would happen if this were repeated with a less good forest mask (say in another location) - how well does the DEFxDIS measure behave in erroneously classified pixels?
L335 - In these plots the DEFxDIS is dominated by DEF, largely because DIS appears less extreme in this location. It might be good to discuss this more, is it a general result? What would happen in areas more dominated by DIS as a primary effect? Is this result due to an anomalous amount of DEF in the field sites leading to this effect being over-stated?
L351 - I think you have buried a headline result here in the table. You have shown here that 67% of Slovakia's forest experienced at least moderate damage (as defined by at least 30% of crown affected) in the heatwave studied. As a headline summary that would be more impactful for use in the abstract etc.
L424-425 - 'Anyway the...' suggest remove 'Anyway the' and start the sentence 'Therefore the contribution...' as this reads as a little too colloquial in English.
Author Response
We thank the reviewer for careful reading of the manuscript and constructive remarks. We have taken all comments into account to improve and clarify the manuscript. Please find in the attachment a detailed point-by-point response to all comments (our replies are in dark blue colour).
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsComments for author File: Comments.pdf
Author Response
We thank the reviewer for careful reading of the manuscript and constructive remarks. We have taken all comments into account to improve and clarify the manuscript. Please find in the attachment a detailed point-by-point response to all comments (our replies are in dark blue colour).
Author Response File: Author Response.pdf
Reviewer 4 Report
Comments and Suggestions for AuthorsThe paper is generally well written, and there are few but important points to be revised
1. In Abstract, Authors should provide quantitative findings (e.g. what is the amount of strong correlation)
2. The Authors stated that they used Level 1C Sentinel 2 products for the analysis, however, the Level 2A products that are already atmospherically corrected are also available for download. In such a study where the spectral responses of the forests are so important, using ground reflectance values (Level 2A) should be preferred. Authors should discuss and necesarily revise on that issues.
3. Authors used one Landsat 9 image but provided very few information about its usage (in which processing level it is obtained, where and how it is used, wihcih bands are selected, how the spatial resolution difference problem is solved in mosaicking and other analysis). Also, I can not see its usage in the flowchart or somewhere else.
4. It is fuzzy that how the classes are determined. I mean how the thresholds are selected to determine 0-10% (or other classess) discoloruation or defoliation or defoltion x discolorization category from their data.
5. As the analysis results are actual maps, it requires class accuracy metrics such as PA, UA, OA, or F1, precision, recall etc. So we can decide if these maps are in an acceptable accuracy. Correlation coefficients or standard error can not answer this question.
6. North arrow is missing in all maps.
Author Response
We thank the reviewer for careful reading of the manuscript and constructive remarks. We have taken all comments into account to improve and clarify the manuscript. Please find in the attachment a detailed point-by-point response to all comments (our replies are in dark blue colour).
Author Response File: Author Response.pdf
Round 2
Reviewer 3 Report
Comments and Suggestions for AuthorsI have no further comments, as the authors have addressed all my concerns and are genuinely trying to improve the draft's quality.
Reviewer 4 Report
Comments and Suggestions for AuthorsAuthors performed intensive revisions and I can see that all of my requests are performed accordingly and answered well.
I only have concerns about the quality of the figure (low dpi or maybe the contrast problem) that may be solved by authors and production team if the paper accepted.