Next Article in Journal
Evaluation of the Impact of Courtyard Layout on Wind Effects on Coastal Traditional Settlements
Next Article in Special Issue
High-Resolution Land Use Land Cover Dataset for Meteorological Modelling—Part 2: ECOCLIMAP-SG-ML an Ensemble Land Cover Map
Previous Article in Journal
Hyperspectral Inversion of Soil Cu Content in Agricultural Land Based on Continuous Wavelet Transform and Stacking Ensemble Learning
Previous Article in Special Issue
Research on Site Selection Planning of Urban Parks Based on POI and Machine Learning—Taking Guangzhou City as an Example
 
 
Article
Peer-Review Record

High-Resolution Land Use Land Cover Dataset for Meteorological Modelling—Part 1: ECOCLIMAP-SG+ an Agreement-Based Dataset

Land 2024, 13(11), 1811; https://doi.org/10.3390/land13111811
by Geoffrey Bessardon 1,*,†, Thomas Rieutord 1,†, Emily Gleeson 1,*, Bolli Pálmason 2 and Sandro Oswald 3
Reviewer 2: Anonymous
Reviewer 3: Anonymous
Land 2024, 13(11), 1811; https://doi.org/10.3390/land13111811
Submission received: 13 September 2024 / Revised: 23 October 2024 / Accepted: 30 October 2024 / Published: 1 November 2024

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Introduction

Lines 25-29: The mention of LULC maps for numerical weather prediction (NWP) models could be more explicit by addressing the importance of these maps in climate modeling and weather forecasting. In other words, explaining the "why" more directly could improve the reader's understanding. Therefore, I believe that reinforcing the relevance of heat fluxes, moisture, and other parameters for weather forecasting, and how LULC maps are important for accurately modeling these fluxes, would be beneficial.

Lines 35-44: The mention of Met Éireann and the HARMONIE-AROME model is specific, but it may seem out of place in the introduction, which should focus on the more general context before diving into specific examples. My suggestion is to simplify the introduction of the Met Éireann example after presenting the broader issue with low-resolution LULC maps and the needs of NWP models.

Lines 52-57: Regarding machine learning, it would be appropriate to highlight the importance of its use. Briefly explaining the advantages of ML in overcoming the limitations of traditional maps could help contextualize the significance of this approach.

 

Materials and Methods

Lines 69-71: It is noted that the term "agreement" is mentioned at the beginning, but its meaning is not entirely clear. I wonder if "agreement" refers to the similarity between maps or to a specific consensus criterion? It would be helpful to add a brief explanation of what "agreement" between the maps means and what criteria are used to assess this concordance.

Results

Lines 333-342: The text mentions specific geographical areas but does not fully explain the relevance of each example for the validation or improvement of ECOSG+. You could emphasize how each selected location illustrates a type of challenge or benefit of the new model. I suggest explaining why these locations were chosen and what they illustrate about the performance of ECOSG+ in different landscapes (urban, rural, remote areas, etc.).

Lines 352-358: The text mentions a "gain in resolution" between ECOSG and ECOSG+, but it would be helpful to explain why this gain is important. The reader could benefit from a more detailed explanation of the impacts of this gain in resolution, such as increased accuracy in areas with small geographical features.

Lines 355-358: The concept of a cut-off value for the quality score could be better explained. The text mentions that areas with lower scores have lower resolution, but it does not explain the impact of this decision on the overall analysis. I suggest that the authors explain the reasoning behind using a cut-off value for the quality scores and how this affects the interpretation of the results.

Lines 367-370: The text mentions that ECOSG+ was downsampled to 300 m to ensure that the differences are not solely due to resolution. The authors could explain in more detail how this downsampling affects the results.

Lines 380-385: The presentation of the F1-score and OA results is clear, but it could be interesting to highlight the improvements in numerical values with direct comparisons, emphasizing where ECOSG+ clearly outperforms the other maps.

Lines 456-458: Add more details about why "Shrubs" is problematic in ECOSG+ and discuss possible approaches to improve the accuracy of this label.

Lines 458-460: A particular curiosity. The absence of identification of labels such as "Boreal needleleaf deciduous forest" and "LCZ5: open midrise buildings" was mentioned. I would like to understand in more detail why these categories are not detected.

Conclusion

Adequate.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper proposes the creation of a 60 m resolution LULC dataset for the continental level. The topic is worthy of research; however, it has several minor changes that need to be addressed before it can be considered for publication.

General comments

C1. The manuscript should include an analysis of regional (and not just global) accuracy, to validate the new land cover dataset, using regional and global reference data.

C2. What was the minimum mapping unit (MMU) and how does it influence the resolution and accuracy of the classification, particularly in urban and rural areas.

C3. In the accuracy assessment, the authors should include the most recent methods from https://doi.org/10.1016/j.rse.2014.02.015 and https://doi.org/10.1016/j.rse.2012.10.031.

C4. A discussion section is needed in which your results are actually compared with those of other researchers. It is necessary to make an exhaustive comparison, I recommend including some relevant references, in order to highlight the novelty of your study, compared to the others.

The answer to these questions should be reflected in the manuscript.

Specific comments

Line 19-30: Authors should describe in more detail the different types of physiographic databases used in numerical climate prediction (NWP) models. Providing specific examples of the parameters that LULC databases help to calculate (such as leaf area index or albedo), may provide more clarity to readers.

Line 32: Change “land-cover classes” to “land-cover classes (thematic resolution)”.

Line 54-56: Why previous methods could not identify the 33 ECOSG labels, and to what extent the new agreement-based approach improves that capability.

Line 59: For the whole world, or a specific region or country?

Line 58-61: What criteria led to the selection of the 60-meter resolution.

Line 120-132: Add concrete examples to illustrate how the specialized agreement score (Ssp) is calculated in specific scenarios. This may improve understanding of how semantic heterogeneity is dealt with.

Line 190: Why was the value of 0.525 chosen, and how does this affect the overall quality of the resulting map?

Line 217-237: It would be useful to include more details on why these specific maps were selected as a reference and if additional accuracy assessments were performed, e.g., why is Dynamic World or Esri Land Cover not included?, see https://doi.org/10.3390/rs14164101  

Figure 3: The authors should focus on land covers. The image should focus on this, improve.

Line 331: How heterogeneity within urban areas is treated is not detailed.

Line 367: The results of the confusion matrix are revealing, but it would be beneficial to provide more discussion on possible causes of confusion between vegetation labels, such as “bare ground” and “crops”. Is this confusion an inherent problem with the input data, or is it due to limitations in the refinement process?

Line 503: The manuscript mentions the need for new specialized maps to improve the representation of problematic labels such as “flooded vegetation”. Consider including suggestions on what type of maps or technologies would be best suited to address this need.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This paper introduces an agreement-based method for producing a 60 m resolution land use/land cover (LULC) map, aiming to enhance the existing ECOCLIMAP dataset's resolution. Overall, the paper is well-structured, with a clear rationale and valuable results. Below are my specific comments:

1. Further Explanation of Methodology: A brief preview of the agreement-based method and how it uniquely contributes to creating the ECOCLIMAP-SG+ dataset would enhance the introduction. This would provide readers with a clearer understanding of the innovation being proposed.The introduction mentions several types of datasets (global, continental, etc.), but it could benefit from a more explicit explanation of how these different categories relate to the research focus on ECOSG labels.

2. By enhancing the intriduction of machine learning methods, the authors can provide readers with a clearer understanding of the approach and its implications for producing high-resolution land cover maps.

3. The expandability of the ECOCLIMAP-SG+ dataset is promising, as its framework can accommodate new data inputs. Future expansions could include:

1) New Datasets: As more high-resolution datasets become available, they can be integrated into the existing framework to enhance the resolution and accuracy of the LULC map.

2) Regional Customization: The methodology can be adapted to create localized versions of the dataset, addressing specific regional characteristics or land use patterns.

3) Improving Accuracy: More comprehensive data can lead to better classification accuracy and reliability in different ecological contexts.

4) Enabling Dynamic Updates: Regular integration of new datasets can keep the ECOCLIMAP-SG+ up-to-date, reflecting changes in land cover and ecological conditions over time.

 

5. Uncertainty Discussion: While the score map indicates heterogeneous quality, a more detailed discussion on the implications of this uncertainty on practical applications would enhance the manuscript.

6. Future Improvements: The authors mention plans to improve the dataset using machine learning. Expanding on the specific machine learning techniques that will be explored could provide valuable insight into future directions.

 

7. Broader Context: While the conclusion focuses on the EURAT domain, a brief discussion on how the methodology could be applied to other geographical regions would enhance the global relevance of the findings.

8. Implications for Future Research: Expanding on the implications of the quality score in machine learning applications could provide readers with a clearer understanding of how ECOCLIMAP-SG+ could be utilized in future research.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Decisão para o Editor

Caro Editor,

O artigo intitulado “High-Resolution Land Use Land Cover Dataset for Meteorological Modelling – Part 1: ECOCLIMAP-SG+ an Agreement-Based Dataset” tem mérito significativo e é adequado para publicação na Land em sua forma atual. As revisões feitas são consistentes com as alterações solicitadas, e a qualidade geral do manuscrito melhorou substancialmente. Parabéns aos autores pelo excelente trabalho.

Back to TopTop