ELULC-10, a 10 m European Land Use and Land Cover Map Using Sentinel and Landsat Data in Google Earth Engine

Round 1

Reviewer 1 Report

The authors submitted a well written and an interesting manuscript dealing with Land Use/Land Cover (LULC) mapping using remote sensing dataset in Google Earth Engine. The methodology is well described and the conclusions are supported by the results. However, the authors need to conduct some revisions before the manuscript could be considered for publication. Below are some comments and suggestions to improve the overall quality of the manuscript.

Lines 45-46: Please provide reference of recent studies conducted using Satellite imagery, Machine Learning (ML) algorithms and cloud computing platforms for accurate large-scale LULC mapping.

Line 64: Please correct “ For instance, [35] exploited Sentinel-2...“ to “For instance, Li at al. [35] exploited Sentinel-2...“

Line 65: Please correct “ [36] also leveraged...“ to “ Ghorbanian et al. [36] also leveraged...“. Please check this citation style for the whole manuscript, please check lines 76, 91, 104...

Line 99: In Table 1, it could be interesting to add the algorithm or classifiers used to generate the available 10-m LULC maps, covering the entire Europe.

Lines 187-196: Since Sentinel-1 satellites acquire data in 3 modes: Interferometric Wide Swath (IW), Extra Wide Swath (EW) and Wave (WV), Please indicate if you used S-1 data collected in Interferometric Wide Swath (IW) mode. Spatial resolution of S-1 is expressed in azimuth and in range (single look), please indicate if this information is available for the dataset you used.

Line 215: The spatial resolution of thermal infrared (band 10–11) bands of Landsat 8 is 100 m. Please correct “30-m thermal infrared (band 10–11) bands“. Please indicate if you used a panchromatic band of 15 m of spatial resolution.

Lines 168-269: in the dataset, please provide some description of ASTER and SRTM dataset presented in the Figure 3.

Line 276: “three types of satellite datasets were employed...“, in the section of Satellite Data Pre-processing, please indicate how these 3 satellites datasets with 3 differents spatial resolutions were pre-processed to reach 10 m spatial resolution of the final product (ELULC 10).

Lines 42-43: Please provide reference of recent studies conducted using remote sensing data, please refer to (1) Muzirafuti, A.; Boualoul, M.; Barreca, G.; Allaoui, A.; Bouikbane, H.; Lanza, S.; Crupi, A.; Randazzo, G. Fusion of Remote Sensing and Applied Geophysics for Sinkholes Identification in Tabular Middle Atlas of Morocco (the Causse of El Hajeb): Impact on the Protection of Water Resource. Resources 2020, 9, 51. https://doi.org/10.3390/resources9040051, (2) Camalan, S.; Cui, K.; Pauca, V.P.; Alqahtani, S.; Silman, M.; Chan, R.; Plemmons, R.J.; Dethier, E.N.; Fernandez, L.E.; Lutz, D.A. Change Detection of Amazonian Alluvial Gold Mining Using Deep Learning and Sentinel-2 Imagery. Remote Sens. 2022, 14, 1746. https://doi.org/10.3390/rs14071746 for water management and  natural resource monitoring.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Dear Authors,

 

Thank you for choosing Remote Sensing to submit your very interesting and anticipated work. I am overall content with the presentation of the results but would like to see some reinforcements with regard to the following - although might be somehow out of the scope of your study:

1. European Union Crop type Map - Acronym: EUCROPMAP - https://data.jrc.ec.europa.eu/collection/id-00346. https://www.sciencedirect.com/science/article/pii/S0034425721004284?via%3Dihub + high resolution layers of Copernicus! They combined cover most of the LULC classes thematically.

2. Chinese GlobeLand30 http://www.globeland30.org/defaults_en.html?type=data&src=/Scripts/map/defaults/En/browse_en.html&head=browse - 2020 edition.

I would also like to see a brief overview in the Introduction  of the EU national GEE maps based on Sentinel-2 / Landsat because such maps exist for some countries. They are also tailored for local conditions and some of them feature high accuracy. If possible, although you compare it in biogeographic regions - you can use their accuracy in the Discussion section too.

 

Kind regards,

Reviewer

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

This study used time sereis of Sentinel-1/-2 and Landsat-8 images to create 10 m European Land Use and Land Cover Map. The authors segmented time-series images and used ANN for classification. The overall workflow is well documented, but the paper structure can be improved. My detailed comments are listed below:

Line 174-175: Suggest switching Figures 1 and 2. Authors can refer to the new Figure 1 here because readers may not be familiar with the seven main biogeographical regions.

Line 198-201: How is the statement related to this study? Which agency provided the data to GEE. What process was done to produce the data?

Line 227: Were survey samples randomly selected or deliberately focused on homogenous areas? Samples in homogenous areas are often easier to classify with higher accuracy. This can significantly affect the validation results and should be discussed.

Line 252: The spatial resolution of points varies from 3m to 40m? How did you integrate the survey data with 10m time series images? 

Section 3.1: If the pre-processings were not conducted by authors, they should be moved to the data description section.

Line 310: Does the product provide surface reflectance directly? It is inconsistent with Line 211, where said TOA was used.

Section 3.4: Why not just add the polygon mask, NDWI, and slope as ANN   input features?

Line 440: Why do Water and two Barren and Artificial Land classes need special post-processing? What about other types of misclassification, such as between water and vegetation in shadowed mountain areas?

Figure 5: Do AUA and APA consider population differences between regions? The APA and AUA of Shrubland seem inconsistent with the results in Figure 9.

Line 509: It would be interesting to compare and discuss the class distributions derived from the map and the reference data.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

Thanks for addressing my comments. I see significant improvements in the revised manuscript, and have no further concerns and suggest to move it forward to publish.