the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
LUCAS Cover photos 2006–2018 over the EU: 874,646 spatially distributed geo-tagged close-up photos with land cover and plant species label
- 1European Commission Joint Research Centre (JRC), Ispra, Italy
- 2Imaging Consulting, Altlussheim, Germany
- 3EFTAS, EFTAS Fernerkundung Technologietransfer GmbH, Münster, Germany
- 4European Commission, Eurostat (ESTAT), Luxembourg, Luxembourg
- 1European Commission Joint Research Centre (JRC), Ispra, Italy
- 2Imaging Consulting, Altlussheim, Germany
- 3EFTAS, EFTAS Fernerkundung Technologietransfer GmbH, Münster, Germany
- 4European Commission, Eurostat (ESTAT), Luxembourg, Luxembourg
Abstract. In the European Union, a tri-annual surveyed sample collects land cover and land use information under the Land Use/Cover Area frame Survey since 2006. A total of 1,351,293 observations at 651,780 unique locations for 106 variables along with 5.4 million landscape and point photos were collected during five LUCAS surveys. In addition to these photos, a set of previously unpublished LUCAS Cover photos were also taken, i.e. following the protocol, a close-up view of the tree, crop, and plant species. These photos contain more details so that tree, crop, and plant species should be identifiable. Between 2006 and 2018, 875,661 LUCAS Cover photos that show the relevant land cover in its entirety were collected. Due to surveyor differences, the images sometimes display elements that require a two-stage deep learning anonymisation process, after which 346 photos were removed before publication. This paper summarizes the collection of LUCAS Cover photos, the filtering for mandatory privacy issues, and provides links to download the data along with the photo metadata, and cross-links to the corresponding LUCAS harmonised survey data. Moreover, after presenting the final public and open dataset consisting in 874,646 photos, potential applications relying on recent advances in geo-spatial analysis and statistical learning such as large scale biodiversity monitoring are discussed.
Raphaël d'Andrimont et al.
Status: final response (author comments only)
-
RC1: 'Comment on essd-2022-117', Anonymous Referee #1, 24 Jun 2022
General Comments:
This paper presents an interesting dataset of geo located photos of land Cover in the EU that was unpublished until now. This dataset is complementary to an existing database of photos from the LUCAS survey. This new dataset (are related paper) is worth to be published.
The abstract and the description of the differences between the photos from the already published dataset and the photos from this new dataset can be improved (see next section)
Technical corrections
The abstract would need to be improved. In particular the abstract cannot be easily understood without reading the manuscript itself. E.g.
- The First sentence in abstract seem to lack a word. Suggestion to add ‘exercise’ as follows: “In the European Union, a tri-annual surveyed sample exercise collects land cover ..”
- The difference between “landscape and point photos” is not obvious (i.e. one has to read the manuscript to get the information)
- Difference between “landscape and point photos” and “LUCAS Cover photos” is not obvious from the abstract alone.
- “LUCAS Cover photos that show the relevant land cover in its entirety” What does it mean? What is the difference or specificity compared to the landscape photos?
- “Due to surveyor differences, the images sometimes display elements that require a two-stage deep learning anonymisation process”. The rationale of the ‘surveyor differences’ is not obvious to understand the need of a ‘ anonymization process’
From section 3 and figure 2 the Difference between landscape photos and “LUCAS Cover photos is quite clear. But the difference between landscape and point photos is still not evident.
-
RC2: 'Comment on essd-2022-117', Anonymous Referee #2, 11 Jul 2022
Geo-located photos are increasingly used in the land cover mapping and biodiversity studies. This study presents a large collection containing 874k geo-referenced photos taken at during the period between 2006-2018. Being different from the previously published LUCAS survery photos, the Cover photo is taken at very close distance to the vegetation so that the leaves, structures of the plants can be clearly recorded. The dataset is undergone through an anonymization filtering and some simple statistics of the data are shown. The dataset is clearly described and the potential applications is discussed. The manuscript can be published after some issues to be addressed.
- The structure of how the photos is organized should be further explained. Currently using the link provided by the authors, under LUCAS_COVER, photos are organized by the years obtained. The subdirectory of the year is country, in the country folder, they are in folders named with three digits number, and in these folders, there are again folders named with three digits number. The file system structure and the meaning of these three digits numbers should be further explained.
- The quality of the photos varies greatly from site to site. The light condition, distance to the target various greatly. For example, LUCAS2012_49204454_Cover.jpg show very dark color and is taken with an upward looking direction. Limited information can be extracted from this photo. Some of the photos are not taken in the field but with samples collected and photos are taken either in a car or a lab (photos in LUCAS2012/FI/492/044). The geo-location of these photos should be checked to make sure that they are correctly labeled.
Detailed comments:
L16, what is the difference between “points” and “locations”?
Figure3. are these cover photos? They are taken close enough to recognize the plant structure.
L86, what are the columns? Are they a from a table?
L112, maybe explain what does xxxx represent in the file name?
Figure4, it would be helpful to add the national boundaries as the background.
Figure5. is it possible to provide more specific classifications (species) of the woodland and shrubland?
Figure6. I do not see photos in categories such as G10, G21, F10, F20, is it because there are less than 300 photos for this class? I would suggest the authors show at least 1 for all classes.
Raphaël d'Andrimont et al.
Data sets
LUCAS Cover 2006-2018 European Commission, Joint Research Centre http://data.europa.eu/89h/c83906d7-1817-495f-b4ba-f2e4b8859d48
Raphaël d'Andrimont et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 268 | 83 | 8 | 359 | 3 | 2 |
- HTML: 268
- PDF: 83
- XML: 8
- Total: 359
- BibTeX: 3
- EndNote: 2
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1