Mapping Dominant Tree Species of German Forests
Round 1
Reviewer 1 Report
The manuscript “Dominant tree species map of German forests” (1751667) develops machine learning based prediction of dominant species across German forest lands.
The manuscript is clear, well written, provides a useful contribution to remote sensing research and merits publication by Remote Sensing. I recommend publication after a few minor additions to the manuscript.
My only concerns center around clarity in some of the methods, and in particular providing some additional information for readers interested in learning or repeating this sort of analyses. Most of this is in the long paragraph starting with line 203.
Page 3 – a small table describing the 10 bands would be helpful. This could even be built into the text as a list, maybe added to the four groups in line 106
Were the 99 (bands) features used in each of the regression models?
At line 205, you say the model was split into regional models. Does that mean 14 separate datasets were trained and tested? Or, was region an added classification feature?
How are the models affected by the steps to differentiate forest/nonforest (207) and broad types (210)? Still 14 models with additional predictors, or now more separate models? For example, were classification type predictive features added to the regional models (e.g., ‘deciduous’)?
Was tuning (line 220) done separately for each of the models?
A small table indicating hyperparameter selection (where different from XGBoost defaults) would be nice.
Similarly, a table of the more important features for final models would be a good addition.
Overall, a very good manuscript with clear presentation of results and an informative discussion.
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
Overall
This study and paper sought to assess the potential of Sentinel-2 data to derive the dominant tree species for forests in Germany. The paper is fairly well written and the primary edit that must be made prior to publication is a much better description of how the accuracies were determined and especially the F1 value. Careful editing for English grammar will improve this paper. Abstract line 8-9 should be corrected for English. "Knowledge of tree species distribution at a national scale is of great importance for a better understanding of the condition of the forests and their potential..." Introduction This section reads well Methods A description of how the accuracy (F1 score) was determined is necessary. Was a standard error matrix created? Later in the results, the reader learns about the outcome but does not know how the F1 score was calculated. Results Beginning on page 10 we see a comparison of NFI and classification (DTS). The F1 values are quite high yet the difference between NFI and DTS is also quite high in some cases. This is problematic and a much better discussion of how all these values were determined is critical. Line 350, was GPS only used? Or was Galileo included? I suspect this should be GNSSAuthor Response
Please see the attachment
Author Response File: Author Response.docx
Reviewer 3 Report
This study presents an approach for the first satellite-based dominant tree species map for Germany. Driven by a time-series of Sentinel-2 data that covers the spectral-phenological characteristics of tree species and sufficient reference data from the NFI, this national map could be generated via a machine learning-based processing pipeline. The dominant tree species map complements the forest information that is available from the NFI sample plots and extends the available forest information in Germany through a full coverage and spatially explicit data set. The paper is interesting but lacks scientific novelty and has a lot of drawbacks:
1. The paper is not written in scientific way, it is more publicistic.
2. The authors should present a flowchart of the proposed machine learning model. What is preprocessing stage of the model? How was the model trained?
3. How was the metrics calculated?
4. Why do the authors use Mann-Whitney-U-Test?
5. The results of the paper are not clearly presented. The authors should compare their results with related papers.
6. The organisation of the paper should be added to introduction.
Author Response
Please see the attachment
Author Response File: Author Response.docx
Reviewer 4 Report
Dominant tree species map of German forests
Mapping dominant tree species on German forests – could this be more suitable?
Line 13 – “NFI data were” - perhaps changing to singular “was”
Keywords: Copernicus – don´t quite agree with this option. German forest, for instance suits better the keywords and work
Line 82 – “IT” – please state the full sentence when using for the 1st time.
Line 99 – on this period were the authors able to compute all the tree annual growth stages to support an accurate pixel classification and NDVI values?
Line 119 – Reference should be given to cloud cover days. Number of days and cloud cover percentage are important information and could be added as impacts the assessments.
Line 233 – were any attempts made to compare and check Sentinel-2 derived data from Corinne Land Cover data? Just to randomly compare the results and having in mind the differences in scales (time and resolution).
Figure 1 – the graph could use the colour scheme from the figure 2. Perhaps a graphical improvement with linked colour will be added-value for readers.
Figure 2 – Forest is normally represented with green colour ramps. Could you please consider updating the image with green pallet? This is evident on figure 3 when forest areas (some in yellow colour –Oak and Beech) are presented side by side with agricultural areas (normally defined by yellow colour). Figure legends must provide the information of the displayed zoomed examples
If you are mapping dominant tree species could you please take some time to improve the maps? Colour schemes need careful look. North arrow needs to be added.
The link to web-based interactive map is not available.
Should readers be able to check it?
The paper seems interesting and provides new insights on the use of remote sensed imagery to map the dominant forest species at Germany national level. Is interesting and stress the importance of forest national inventories and points to the lack of high resolution forest data.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 3 Report
The paper is still more publicistic and its scientific novelty is extremely doubtful. The authors miss a lot of technical details. Figure 2 is unclear. Metrics are unclear.
