the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Map of forest tree species for Poland based on Sentinel-2 data
Abstract. Accurate information on forest tree species composition is vital for various scientific applications, as well as for forest inventory and management purposes. Country-wide, detailed species maps are a valuable resource for environmental management, conservation, research, and planning. Here, we performed the classification of 16 dominant tree species/genera in Poland using time series of Sentinel-2 imagery. To generate comprehensive spectral-temporal information, we created Sentinel-2 seasonal aggregations known as Spectral-Temporal Metrics (STMs) within Google Earth Engine (GEE). STMs were computed for short periods of 15–30 days during spring, summer, and autumn, covering multi-annual observations from years 2018 to 2021. The Polish Forest Data Bank served as reference data, and, to obtain robust samples with pure stands only, it was validated through automated and visual inspection based on very high resolution orthoimagery, resulting in 4500 polygons, serving as training and test data. The forest mask was derived from available land cover datasets in GEE, namely ESA World Cover and Dynamic World. Additionally, we incorporated various topographic and climatic variables from GEE to enhance classification accuracy. The Random Forest algorithm was employed for the classification process, and an area-adjusted accuracy assessment was conducted through cross-validation and test datasets. The results demonstrate that the country-wide forest stand species mapping achieved an accuracy exceeding 80 %, however it varies greatly depending on species, region and observation frequency. We provide freely accessible resources including the forest tree species map, training and test data: https://doi.org/10.5281/zenodo.10180469 (Grabska-Szwagrzyk, 2023).
- Preprint
(1639 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on essd-2023-482', Oleksandr Melnyk, 31 Jan 2024
Dear authors,
In general, the article made a pleasant impression, is well written and fully represents the methodology and results of the study.In this kind of research, the most important challenge is to create training samples. Based on open FDB data and to improve classification accuracy, it is worth conducting field validation, but on the scale of even regions, this task is very difficult and time-consuming. On the other hand, the frequency of FDB data updates is important. The classifier's accuracy may be impaired by deforestation that is overgrown with fast-growing vegetation in a year or two, which we have encountered in our research.
It is not clear from the text of the article why the Random Forest classification algorithm was chosen. This algorithm is very popular among researchers, although there are other effective algorithms whose results would be interesting to compare, especially on a national scale.
Perhaps, to improve the quality, it is worth dividing the territory of Poland, for example, by geographical provinces, although such a division is rather arbitrary.
The suggestions I have made do not in any way affect the quality of the work, and therefore I recommend it for publication.
Citation: https://doi.org/10.5194/essd-2023-482-RC1 -
RC2: 'Comment on essd-2023-482', Jan Hemmerling, 09 Feb 2024
Dear Authors,
this manuscript covers a topic of constant relevance, is clearly written and for the most part easy to follow. The results are also sensibly discussed and summarised.
However, there are still some questions that need to be addressed with regard to the methods. For the reader, it is not clear how the multi-year aggregated features were created; in some places STMs are mentioned (e.g. lines 136, 138, 141), in others composites (lines 142, 147, 150). If these are composites, it would be important to know which rule set was used to create them. In the case of temporal statistics, it would be important for traceability to know which metrics were included as features in the classification. A list/overview of all features included in the classification would also be desirable. This also applies to the additional explanatory variables (from line 167). Has the effect of these variables on the classification results been tested?
The period of the time windows for the creation of the composites/stms seems to overlap, at least in 2021 (Table 2.). Why was the summer period not shifted back? Is it possible that the same information was received in both STMs/composites?
To compensate for the differences in the number of training data, additional training data was taken from less common tree species from stands with a 60-80% mixing ratio (line 105). What proportion of the total number of trainingpixel in these classes did this account for? I think this is definitely relevant for the interpretation of the results. Perhaps this could be added to the appendix?
In general, however, I think that once these aspects have been clarified, nothing stands in the way of publication and I look forward to receiving comments on my remarks.
Citation: https://doi.org/10.5194/essd-2023-482-RC2 -
EC1: 'Comment on essd-2023-482', Nophea Sasaki, 29 Feb 2024
Dear Authors,
Thank you for submitting your manuscript to our journal. We have received feedback from two reviewers, who have assessed your work. Both reviewers acknowledge the importance and relevance of your research, praising the clarity of writing, the comprehensive discussion of results, and the methodology employed in your study. However, there are several aspects that need further clarification and revision to enhance the quality and comprehensiveness of your paper.
Reviewer #1 Highlights:
- The justification for choosing the Random Forest classification algorithm over other algorithms.
- The potential improvement of classification accuracy through field validation and consideration of FDB data updates frequency.
- A suggestion to divide the territory of Poland by geographical provinces to possibly improve the study's quality.
Reviewer #2 Highlights:
- Clarification on the creation of multi-year aggregated features, including the rule set for composites and metrics for temporal statistics.
- An overview of all features included in the classification and the effect of additional explanatory variables on classification results.
- Questions regarding the overlap of time windows for composites/STMs creation and the proportion of training data from less common tree species.
Both reviewers recommend the publication of your manuscript once these issues are addressed. Therefore, we invite you to revise your manuscript based on the reviewers' feedback and resubmit it for another round of review. Please ensure that your revised manuscript includes a detailed response to each point raised by the reviewers, highlighting the changes made to the manuscript.
We believe that addressing these comments will significantly improve the manuscript and look forward to your revised submission. Should you have any questions or need further clarification, please do not hesitate to contact us.
Thank you for considering our journal for your work.
Citation: https://doi.org/10.5194/essd-2023-482-EC1 - AC1: 'Comment on essd-2023-482', Ewa Grabska-Szwagrzyk, 20 Mar 2024
Data sets
National-scale tree species/genera map for Poland from Sentinel-2 time series Ewa Grabska-Szwagrzyk https://zenodo.org/records/10180469
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
478 | 115 | 25 | 618 | 22 | 22 |
- HTML: 478
- PDF: 115
- XML: 25
- Total: 618
- BibTeX: 22
- EndNote: 22
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1