| 10 Mar 2026
Lessons learned from UAV surveys over drifting sea ice at high latitudes
Abstract. In this study, we provide an overview of the challenges encountered when operating consumer grade Unmanned Aerial Vehicles (UAVs) or drones over high-latitude Arctic pack ice. A major obstacle in generating void-free orthomosaics and Digital Elevation Models (DEMs) of ice floes is their drift speed, which can easily exceed 0.5 m s-1 in the central Arctic Ocean. Furthermore, there is no commercial solution for operating UAV grid flight missions north of 85° N latitude. To address these issues, we tested and present a novel approach for flying autonomous grid missions at high latitudes including a pre- and post-flight drift correction scheme. Using the resulting data, we generated drift-corrected orthomosaics and DEMs for all ice floes visited during the ArcWatch-2 cruise in August/September 2024. In addition to the UAV camera system we tested a drone based laser scanner and incorporated measurements from drifting Global Navigation Satellite System (GNSS) stations for referencing and validation. Finally, we demonstrate the potential of drone acquired data for deriving ice thickness estimates, surface type classifications and melt pond depth retrievals. Our results offer the potential to upscale in situ measurements and can contribute to better mass balance estimates and meltwater quantification.
In this study, the authors present the method that they have applied to map the sea ice surface properties using UAV-based photogrammetry and LiDAR measurements. This work represents a very good and much needed reference for the growing UAV community that intends to use drones in the extremely challenging Arctic environment. To improve the quality and usability of the paper, I invite the authors to address the following three main points, as well as the detailed comments below:
I invite the editor to consider the acceptance of the paper after the authors have addressed these concerns.
Detailed comments:
Line 17: comma is missing after “trivial”
“While flying off-the shelf drones over stable ground at mid latitudes is almost trivial, we encountered several problems
Line 59: “along track and across track respectively results…” should be “along and across track, respectively, results…”
Line 63: comma is missing after “ocean currents”
Line 64: comma is missing after “for this”
Line 65: comma is missing after “Otherwise” and after “guaranteed”.
Line 66: remove “single”
Line 67: comma is missing after “prior to the flight”
Line 68: “pseudo Lagrangian grid mission”: I think that calling the grid “Lagrangian” or “pseudo Lagrangian” is not appropriate. A Lagrangian perspective would follow the motion of a target in time, while here the time dimension is absent. Although the photos of the orthomosaic are captured at different times, they do not include any temporal variability of the targeted ice surface, the whole orthomosaic represents a snapshot of the ice condition at a single time. Accounting for the ice drift in the flight pattern (Fig 2a) does not make the pattern “Lagrangian”, and removing the drift (Fig 2b) does not make the pattern “Eulerian”. A time dimension would be needed to invoke the “Lagrangian” or Eulerian” flows.
Fig. 2 caption: comma is missing after “mission planning” and “post-processing”. The comma after (a) should become a full stop, the bracket after “arrows” should be removed, and a bracket before b) should be added. The last sentence is a bit too cryptic and hard to read: to make it clearer, I would replace “GNSS drift estimates” with “the ice drifts estimated from GNSS sensors”, and I would avoid introducing the concept of “Eulerian coordinate system”, which sounds ambiguous in this context (see comment above). Instead, I would precisely explain what you did, i.e. you translated the image locations to the positions in which the photographed ice was at fixed reference times.
Line 73: comma is missing after “Before takeoff”
Line 79: comma is missing after “quality check”
Line 81: comma is missing after "flight routine” and after “For this”. Add “-“ in “drift-corrected”.
Line 82: replace “exif metadata” with “EXIF (Exchangeable Image File Format) metadata”
Line 87: throughout the whole paper, please add “-“ between “drift” and ”corrected”
Line 88: “GNSS solutions” us unnecessarily vague: do you mean
“Additionally, the GNSS solutions from our drifting reference stations at the reference times listed in Table 1 were imported into the software and were used as GCPs during the processing”, To me, this sentence seems quite tortuous. To simplify it and improve clarity, I suggest to rephrase to something like “as well as the GNSS-base GCPs at the reference times listed in Table 1” and replace the full stop of the previous sentence with a comma. In fact, in Sect 2.1 you already explained that the GNSS stations were used as GCPs.
Line 90: add a comma after “images”
Line 92: to improve the language, I would replace “now aided by the more precise GNSS reference solutions.” with “using the GCPs”.
Line 93: add a comma adter “DEMs”
Line 94: please replace “next” with “successive” and add commas after “step” and after “generated”.
Line 95: add a comma after “DEM”
Line 96: To improve clarity, I would replace “we applied an additional fit to the data to adjust with sea level” with “we applied an additional fit to detrend the data” It would be good to add here some reference to previous work where this detrending fit was applied to DEMs.
Line 97: add a comma after “step”.
Line 107: replace “the pond level” with ”the height of the pond”. Please use past tense instead of present tense in the last three lines of Sect 2.2.3.
Line 108: “we derive pond depth in a two-media photogrammetry approach”: could you please add a brief explanation of the applied approach, so that the reader does not need to read Fuchs et al. (2024) to understand the method? That should also clarify what the “brightness correction” mentioned in Fig 3a refers to.
Line 118: replace “a nearby” with “the” (the mobile RTK station was described in the section above so it is not a generic device). Please use past tense in the sentence.
Line 120: please explain what “LAS” stands for.
Line 125: add a comma after “DEMs”.
Line 126: please replace “DEMs” with “LiDAR-based DEMs” to make clearer the distinction between the two DEMs
Line 128: add a comma after “Arcwatch-2”
Line 134-135: “In the next step both datasets were resampled onto a common, equidistant track at 1.5 m point spacing so that each point contains both snow and ice-thickness information”. This sentence is not very clear: did you mean “both datasets were resampled onto a common track of equidistant measurement points having 1.5 m spacing, so that …” Please also remove “In the next step” and use past tense instead of present tense.
Line 136-138: “The clearly visible footpaths in the corresponding orthomosaics allowed for additional horizontal corrections by aligning the equidistant track files to these features.” Also this sentence is not clear: do you mean that you used the GEM tracks as reference distance in the processing of the UAV photos to generate orthomosaics and DEMs? Did you align files or the data points that belonged to the GEM track nd that were visible in the UAV photos? Please rewrite the sentence to clarify it.
Line 139-144: This paragraph does not belong to this section. It should be moved to the Results section, merging it with the sentence in line 159-160. Please explain the sentence: “For this we converted the ellipsoidal DEM elevations into freeboard estimates employing the in situ freeboard measurements conducted at each GNSS station.” How did you use the in situ freeboard measurements? Did you use them as vertical reference distances in the processing of the UAV photos to generate the DEMs? If so, please state it explicitly. What is the estimated uncertainty of the freeboard derived from DEMs in places far away from the GNSS stations?
Table 2. The table caption does not clarify the content of Table 2. In the sentence “Here we show the average deviation to all GCPs with No GCPs, 3 GCPs, All GCPs and if required with an Extra fit included in the photogrammetric processing.” what “average deviation” refers to? Is the mean horizontal distance between the GCPs identified in the orthomosaics and the measured GCPs location? If so, are the numbers given in the table “horizontal distances +- vertical uncertainties”? Mixing horizontal and vertical distances/uncertainties does not make much sense to me… probably I misunderstood. Can you clarify? Also, could you explain in the text how you calculated the vertical uncertainty (1sigma)? The measured freeboards at the GNSS stations were used to generate the DEMs, so those measurements could not be used for validation. What did you use then for validation?
Line 147-148: “Final DEMs include all GCPs in the processing and were adjusted with an additional vertical fit to adjust with sea level if required.” Are you here referring to the detrending fit described at the end of Sect 2.2.2? If this is the case, please remove this sentence as it is a repetition, and instead I recommend to better explain what the “Extra fit” in the last column of Table 2 is, in the caption of Table 2.
Line 176-177: “Even under poor environmental conditions data from grid flights can be used for stitching orthomosaics making it an ideal tool for quickly generating orthorectified maps of large ice floes.”: please clarify which conditions you are referring to, and please add a comma after “conditions” and after “orthomosaics”. As it is now, this sentence does not sound very meaningful to me. I guess that you mean that grid flights become a doable option even over drifting ice only if the suggested pre-flight and post-flight drift corrections are applied? Please clarify.
Line 190: “Also the wavelength of 905 nm is perfectly suited for snow and ice applications.” How is this conclusion reached? Is this the wavelength of the applied LiDAR? If so, this should be mentioned earlier in the method section. Since no other wavelength has been tested in the paper, the authors should explain that the comparison with GEM observations suggested that the wavelength used in the LiDAR measurements allows retrievals of ice thickness with an accuracy of …
Line 197-199: “We assume that the ice at the pond bottom of the here observed melt pond has potentially lost scatterers, such as air inclusions through an increased permeability, and thus becomes more translucent, causing a reconstruction of the sub-pond bottom layer.” This sentence is totally obscure because the method applied to retrieve the pond depth from RGB photogrammetry is not explained. And what is the “sub-pond bottom layer”?
Line 200: “Another small part…” I invite the authors to use more proper language such “Other sources of error may have contributed to the observed discrepancy between measured and retrieved pond depth, such as…” or similar phrases.
Line 205: “… error dependencies on pond evolution and morphology, and the technology used”. The language used in this sentence should be improved. Does the meaning here is that future campaigns will enable a better validation of the retrieval method, and a better understanding of its limitations in different stages of the pond evolution? Please explain.