the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 05 Feb 2026
Brief communication: Improving botanical monitoring in proglacial areas with high-resolution UAV data
Abstract. Proglacial areas are undergoing rapid ecological and land cover changes as glaciers retreat globally. Field campaigns are essential in proglacial environments to monitor the colonization of the deglaciated areas by the vegetation. In this study, we evaluate the benefit of using optical UAV images to optimize botanical data collection during these campaigns. We compared the vegetation cover measured in situ by experts with the vegetation cover derived from high-resolution UAV images in a high-mountain environment below the Aneto glacier in the Pyrenees. Retrieving the vegetation cover from UAV images provided valuable and complementary data to traditional in situ detailed observations.
- Preprint
(930 KB) - Metadata XML
-
Supplement
(543 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-527', Jan Komarek, 18 Feb 2026
-
RC2: 'Comment on egusphere-2026-527', Anonymous Referee #2, 04 Mar 2026
Dear authors,
overall the topic is relevant and interesting as glaciers are receding faster and thus also the enlargement of the proglacial areas.
In general the abstract is very vague, it would nice to have a short information about the result.
As you are working in a proglacial area it would be more important to cluster the plots by years ice-free than elevation and also the whole proglacial are should be clustered by years ice-free, then it would be better to compare with the adjacent area outside.
Introduction:
Line 20: a comma is missing between local topography, and soil properties
Line 29-30: However, in situ observations fo the vegetation are only possible at a few dozens of plots of typically 1 m² to 4 m² - this is not totally correct as other studies use 10 m² plots (5x2 m) and estimate cover in each 1 m² subplot - (see among others: https://doi.org/10.5772/intechopen.69479, doi:10.3390/d12050191)
Methods:
Regarding the methods it is not clear how the cover values from the UAV images were derived. You refer to Deschamps-Berger et al. 2020, but this reference is missing.
It is also not clear how control points were choosen and if they were marked in the field.
Could you also give information in the text what was the minimum cover estimated in the field.
What would have been helpfull for getting cover values and classify pixels in the UAV images as vegetation, would have been the calculation of vegetation indices like the NDVI but derived from RGB images - there are a lot of such indices available.
Results:
the points available for comparison differ between methods and results
Fig. 3: it is not clearly showing the differences between in-situ measurements and cover derived from UAV image classification, this would be important because you want to evaluate the UAV cover by using in situ data
Fig. S2: Please highlight the large plants because one can identify them only when zooming in in the PDF on the computer, then it would be much clearer for the reader.
Citation: https://doi.org/10.5194/egusphere-2026-527-RC2 -
RC3: 'Comment on egusphere-2026-527', Anonymous Referee #3, 04 Mar 2026
I carefully read the manuscript “Brief communication: Improving botanical monitoring in proglacial areas with high-resolution UAV data.” The manuscript is well prepared and clearly presents the research questions and the workflow adopted to address them.However, the abstract is quite brief and should include some information about the main results. Since the study performs an automatic classification of several land cover classes, it would be useful to provide a confusion matrix to better evaluate the classification performance. I would also like to ask the authors for a small clarification regarding the use of optical RGB data. Could the introduction of near-infrared (NIR) data potentially improve the performance of classifications at lower spatial resolutions? Vegetation is highly sensitive to this wavelength, and many vegetation indices rely on NIR. A short discussion of this aspect could strengthen the manuscript.
I also suggest a few minor improvements to the figures:
Figure 1: I suggest adding the dates of the glacier fronts in the legend.
Figure 2 (vegetation cover plot): I suggest adding borders to the symbols to better distinguish the points from area A1 from those belonging to areas A4–A6.
Overall, the article can be accepted after these minor revisions.
Citation: https://doi.org/10.5194/egusphere-2026-527-RC3
Data sets
UAV land cover in the proglacial area near the Aneto Glacier César Deschamps-Berger, Jesús Revuelto, Francisco Rojas Heredia https://doi.org/10.5281/zenodo.17989880
Model code and software
aneto_glacier_vegetation César Deschamps-Berger https://framagit.org/cesardb/aneto_glacier_vegetation/-/tree/publication_UAV_article
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 148 | 71 | 25 | 244 | 43 | 51 | 50 |
- HTML: 148
- PDF: 71
- XML: 25
- Total: 244
- Supplement: 43
- BibTeX: 51
- EndNote: 50
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Dear authors,
Your paper addresses a relevant and timely topic of optimising botanical monitoring in rapidly changing proglacial environments. However, in its current form, the manuscript lacks the technical depth and thematic clarity required to bridge the gap between remote sensing and field ecology.
The abstract is very general and does not present results. The conclusion here is vague. After reading, there is no need to read the paper. Sell your paper and attract readers' attention.
The introduction should also be more specific. RS (drones included) has supported vegetation mapping for years. What are the gaps? There is a known misunderstanding between ecological mapping and RS surveys based on thematic resolution, etc. I would appreciate it if you were more specific about the gaps and potential methods. All three RQs are vague, could be answered immediately. Read state-of-the-art and name specific gaps. Which vegetation cover did you map? Why such? Which strategy did you use for bridging in-situ and RS?
Methods. The study area is impressive. But I miss any information about vegetation cover. Perhaps you should introduce typical species (and their common sizes, as these are probably crucial for understanding your results).
I also wonder about the 3rd polynomial georeferencing. Any reason for doing this? Do you have a quantitative control of georeferencing? The systematic shift caused by GNSS should be treated as a 1st-order at maximum at most once the imagery is aligned using SfM. Why was some land cover masked manually, and others using RF?
Results. The results show a mixture of RF potential misclassification, with altitude as a confounding factor. The results indicate that a 10+ cm plant can be identified accurately using sub-centimetre imagery.
“Only two plots were available in the 1 cm map, which did not allow for robust statistical calculations.” So why did you acquire them?
Figure 2 suggests that only 0.1 cm makes sense, while others underestimate. Well, perhaps it is connected with MMU, which probably should be defined at the very beginning, together with the botanists.
Figure 3 presents the expected decrease in vegetation cover. It seems there is no overlap between in-situ and UAV data at A4-A6. Consider changing the whole chart. It would also be nice to present a table with a quantitative evaluation.
Discussion. You conclude that the higher elevation would require higher spatial resolution UAV imagery. Without information about specific species and their size, it's kind of an empty proclamation. A figure from the ground may help readers to better understand. It is a bit difficult to understand the environment in which 0.1 cm is insufficient for vegetation cover detection. Sub-centimetre resolution is no longer a mixel.
The shadow effect is not discussed at all. What was the proportion of shaded vegetation? How did this influence the results? That could mention bias.
The in-situ usually includes mosses and lichens, which tend to be spectrally indistinguishable from the rocks or shadows. This creates a thematic mismatch. The reported underestimation may be due to the detection of vascular plants alone. This may introduce another bias.
Conclusion. I would not claim that “UAV vegetation cover showed clear bias compared to in-situ”. You need to deal with or discuss the bias first. There is obvious disagreement over what is considered vegetation by botanists and RS. And it was not explained, actually.
All the best,
Jan Komarek