Observational data of Arctic Sea Ice Melt Ponds: a Systematic Review of Acquisition and Processing Approaches

Aparício, Sara; Driscoll, Simon; Flocco, Daniela

doi:10.5194/egusphere-2025-4480

Preprints

https://doi.org/10.5194/egusphere-2025-4480

Preprints

09 Oct 2025

| 09 Oct 2025

Observational data of Arctic Sea Ice Melt Ponds: a Systematic Review of Acquisition and Processing Approaches

Sara Aparício, Simon Driscoll, and Daniela Flocco

Abstract. This review synthesizes current methods for acquiring and processing Earth observation (EO) data relevant to Arctic sea ice melt ponds (MPs), pools of meltwater that form on the ice surface during the polar summer. By reducing albedo, MPs amplify the ice–albedo feedback and alter the sea ice energy budget, exerting a strong influence on the Arctic climate system. Robust observational records are therefore essential for improving sea ice prediction in a rapidly changing and highly sensitive polar environment. Despite this importance, melt pond parameterizations remain underdeveloped in many sea ice models. Advancing these parameterizations, through refinement of existing schemes and integration of novel approaches, is a critical priority for better constraining sea ice evolution and its role in the climate system.

Here we review the main EO methods used in MP studies, including active and passive optical sensors (multispectral and LiDAR) and microwave instruments (synthetic aperture radar, radiometers, and scatterometers). We also summarize melt pond signatures across the electromagnetic spectrum, outlining the strengths and limitations of each sensor. Complementary in situ observations from field campaigns, together with key processing techniques, are discussed, alongside a synthesis of available MP datasets from satellite missions and ground-based campaigns. Persistent EO data gaps, such as cloud cover, limited temporal sampling, and spatial constraints that lead to underrepresentation of different Arctic regions and ice types, remain a major challenge, highlighting the need for future missions with improved resolution, coverage, and spectral capacity.

By compiling and critically assessing these datasets and methods, and identifying current knowledge gaps, this paper provides the most comprehensive review of melt pond observations currently available. It is designed to support refinement of parameterizations and the development of multi-modal modelling approaches, crucial for addressing observational gaps and ultimately advancing the understanding and prediction of Arctic change.

Received: 15 Sep 2025 – Discussion started: 09 Oct 2025

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Sara Aparício, Simon Driscoll, and Daniela Flocco

Status: final response (author comments only)

Subscribe to comment alert

RC1:
'Comment on egusphere-2025-4480 (Review paper)', Anonymous Referee #1, 24 Oct 2025
This manuscript aims to provide a comprehensive review of observational approaches for quantifying melt ponds on Arctic sea ice. Given the central role of melt ponds in ice-albedo feedbacks and the difficulty of obtaining consistent observations, a systematic overview comparing the strengths, weaknesses, and future potential of various methods is highly relevant and well-motivated.
However, the current version of the manuscript lacks the focus and structure expected for a review paper. The organization makes it difficult to follow the line of reasoning between sections, and the conclusions remain unclear. The manuscript should better clarify the current state of sensing capabilities and offer practical guidance on which products or methods are best suited for different applications.
Major Comments
Organization and Focus

The manuscript is often repetitive, particularly in sections discussing general motivation for melt pond observations. In contrast, the technical comparison of methods lacks sufficient depth.

Section 3 should begin with a concise synthesis of overarching challenges common across observational platforms, including:

Temporal coverage (limitations due to clouds, daylight, and repeat time),

Spatial resolution, and

Discrimination and identification of pond boundaries.

These challenges are especially critical for melt pond fraction (MPF) retrievals.

Scope and Content

It is unclear what the extent of observational datasets that the manuscript aims to include are. The section describing field campaigns should clarify which are included and why. Focusing on those most relevant for benchmarking or validation would strengthen the section. A map of campaign locations could help illustrate spatial biases.

Although the focus is on Arctic sea ice, some mention of Antarctic melt ponds is warranted. A short discussion of their sparse observations, unique challenges, and potential for future monitoring would improve completeness.

The section on parameterization in global climate models seems beyond the scope of this review. It could instead be reframed as part of the motivation or discussion of future directions.

Key Variables and Definitions

Introduce the key melt pond variables early in the manuscript, ideally in a table. These should include MPF, depth or volume, connectivity (and open vs. lidded), and melt onset.

Methods for measuring parameters beyond MPF (e.g., pond depth or volume) need greater attention (see Buckley et al., 2023; Fuchs et al., 2024).

Figures and Tables

The choice of figures should be revisited. A meta-analysis figure summarizing validation efforts or comparisons of MPF products would be valuable. Notably, a figure showing typical seasonal cycles should be included.

Consider adding:

A consolidated table summarizing pros and cons of each remote sensing method for MPF (the current figure does not allow easy comparison).

Example images comparing different types of retrieved melt pond products.

A space–time diagram quantifying temporal and spatial coverage, or the percentage of usable data for each platform.

Specific Line-by-Line Comments
L52: Pond color has been shown not to strongly depend on pond depth. Revise or list last.

L66–67: The statement that melt ponds can break the spring predictability barrier lacks clear supporting evidence.

L68, L70: Reconsider the choice of references (e.g., Driscoll et al., 2024; Polashenski et al., 2012).

Figure 2: If retained, the text should better explain why this observation is relevant.

L123–124: Repetitive—streamline.

L137–142: Provide references for albedo values and ensure consistency across lines.

Figure 3: Consider annotating observable variables and how features such as refreezing or lidding affect retrieval.

L155–157: Add references.

L172: Typo—merge sentences.

Figure 4: Well prepared, but note limitations (e.g., LiDAR also requires cloud-free conditions). Clarify whether each method detects or times onset.

L255: Replace “capabilities” with “utility.”

L277: Clarify whether “liquid melt pond fraction” differs from MPF used elsewhere; ensure consistent terminology.

L320: Clarify whether sensor sensitivity differs for FYI vs. MYI, and introduce this distinction earlier.

L366: “Firn” is not relevant for sea ice melt ponds—remove.

L387: Clarify intent and scope of listed observational campaigns.

Figure 6: Consider removing or supplement with date, location, and retrieved pond parameters.

L425–434: Reorganize to emphasize what was observed, derived products, and how data were collected.

L468, L474: Minor wording edits (“melt pond”; “specifically”).

L480–484: Redundant—streamline.

L484: Consider starting a new section; summarize datasets and melt pond variables.

Section 3.3: Difficult to connect with sensor overview. Suggest merging with Section 4 and organizing by wavelength/sensor type.

L526: Replace “convergent” with “converged.”

L538: Table 4 missing—add or renumber.

L547–549, L553: Typographical and introduction issues—revise.

Table 2: Clarify whether “MERIS-ZEGE” is identical to MPD1; include sensor type for consistency.

Section 4.2: Scope too broad—consider omitting or condensing to future directions.

L666–697: This list adds little—remove or condense.

L671: Incomplete sentence—revise.

L714: Clarify meaning of “sensor-based constraints.”

L734: Highlight the importance of seasonal transitions more clearly.

Section 5.2: Possibly redundant—merge with earlier priorities.

L794–800: Serves as general motivation, not melt pond–specific—condense.

L801 onward: Repetitive—edit for conciseness.

References
Buckley, E. M. et al. (2023). Observing the evolution of summer melt on multiyear sea ice with ICESat-2 and Sentinel-2. The Cryosphere, 17(9), 3695–3719.
Fuchs, N. et al. (2024). Sea ice melt pond bathymetry reconstructed from aerial photographs using photogrammetry: a new method applied to MOSAiC data. The Cryosphere, 18(7), 2991–3015.
Citation: https://doi.org/10.5194/egusphere-2025-4480-RC1
RC2: 'Comment on egusphere-2025-4480', Anonymous Referee #2, 13 Nov 2025

Please refer to the attached document.

Citation: https://doi.org/10.5194/egusphere-2025-4480-RC2
CC1: 'Comment on egusphere-2025-4480', Stefan Kern, 17 Nov 2025

Comment to
Observational data of Arctic sea ice melt Ponds: a systematic review of aquisition and processing approaches
by Aparício, S., et al.
General Comment: This is an impressive piece of work and I am glad that it is already online as a citable discussion paper. I went through the manuscript - not necessarily with a reviewer's eye but with the intention to learn something new and to provide ideas how readability and usability of this publication could eventually be improved further. Perhaps one or more comments could assist here.
L1-2: This is a very high-level title which, at least from my side, raises very high expectations. While you manage to come up with an impressive compilation of various data sets and also attempt to describe sensing and processing techniques I suggest to come up with an alternative structure to guide the reader a bit better.

I was wondering whether it would make sense to begin - as you did - with the relevance of melt ponds for studying the Arctic climate and selected (!) studies where observations of melt ponds changed and/or influences our knowledge about Arctic summer sea ice conditions. I think in that section if would not matter whether you are referring to satellite, airborne or ground-based studies - simply because all have their different application areas and examples.

Then, I would come up with the section where you describe the 3 main different observational tools: ground-based, airborne, spaceborne in 3 sub-sections. In each of these you would refer to the measurement technique, provide a list of the sensors and their characteristics, experiments / expeditions / satellites and at the end of each of these 3 subsections come up with the limitations of use / knowledge gaps / and room for improvement.

I would try to provide tables and/or appendices that are clearly linked to these 3 sub-sections, i.e. in-situ, air-borne and satellite.

I don't think it makes a lot of sense to try to distinguish between pan-Arctic and regional spaceborne applications because in the long run, applications such as from Sentinel-2 MSI might become pan-Arctic as well once there is enough coverage. I guess, if described properly, readers will understand that 10 km x 10 km large super-high resolution satellite images are not pan-ARctic and are not suitable for climate studies but - like the air-borne data sets - are perfect for algorithm development and evaluation.
Figure 4: I think the row detailing the signature would need much more information because currently it mixes information about signature changes during onset and mature melt pond existence. I find this illustration overly simplistic and partly also misleading. I can provide details if needed.
L196: You could add Lee et al., 2020 as well.
L249-252: Clouds are an issue as well. And: ICESat-2 offers a very limited daily coverage - hence major steps of the seasonal development of the melt pond evolution might remain undetected.
L254-256: This statement is not entirely correct because water vapor influences frequencies near 22 GHz and the higher (near-90GHz) frequencies are increasingly influenced by the atmosphere (water vapor and clouds).

All frequencies are in some way or another influenced by wind-induced roughening of the water surface and, for the lower frequencies, also by the water salinity.
L295++: The list of key challenges looks good ... but, even though Fig. 5 is relatively clear about the interaction properties between the radar signal and the medium, I strongly suggest to be more specific when you talk about penetration depth (only sea ice and/or snow - but also merely when both are dry, but not open water. Melt pond depth has no influence radar backscatter). Also, when you mention seasonality you should stress better what you mean by that. The "seasonality" of the melt pond plays no role but the seasonal change of the physical properties relevant for its microwave remote sensing between the melt ponds. As a rule of thumb I would say: wind-roughened and hence "bright" melt ponds can be detected in smooth sea ice (landfast sea ice) while smooth melt ponds (no wind) can be detected in rough (deformed) sea ice - as the two best resolved special cases. Everything inbetween is a bit "wishy-washy".
L360-369: In this paragraph findings and physics are mixed between land surfaces and sea ice. I suggest to concentrate on sea ice because melt onset detection on land (and ice sheets) is different from that on sea ice.

Also I am not so sure whether you should not mention all sensors first (but forget about SMMR) and then write about the potential application and studies carried out.
L378-380: There might be more recent literature discussing the uncertainties melt ponds can cause in sea ice concentration retrieval (you could check The Cryosphere 2016 and 2020). The 2nd sentence needs reference to melt-pond fraction Retrieval I guess.
L489/490: This link does not work.
L500: You are refering to Appendix A and also provide a few examples of approaches in this paragraph. But you do not link back to the descriptions you have compiled earlier on. Here I see room for structural improvement.
L501: "Early approaches" --> not sure I understand what you mean by that when citing papers of years 2015+.
L505: Kern et al. 2016 did not develop any such approach. They used the data. Possibly you wanted to refer to Tschudi et al., 2008 and Rösel et al. 2012.
L515-527 ... this goes back to the comment I made in the context of line 500 ... I would find it more logically to combine methods and data sources, i.e. satellite sensors in one go. I am sure when constructing this manuscript the order of topics was made on a conscious decision but I miss to understand why.
L563/564: As far as I know there has not been any systematic and independent study intercomparing the various melt-pond fraction data sets. Therefore I would be very careful in making statements why one data set or another is or was considered better (or with the higher) accuracy than others. Often data set producers evaluate their products with a biased view, understandable because they want to "push" their data set upfront. The two studies (Lee et al. and Ding et al.) accordingly highlight the performance of the respective (their) products. But the truth is that all the data sets (particularly the pan-Arctic ones) have their deficiencies and/or are heavily based on more or less ideal assumptions.
L602: I suggest to refer to the data sets by the author team name - unless there has really been an expedition such as in case of HOTRAX.
L602-615: I note that here you mix satellite, airborne, and ground-based observations ... I find this not to be an optimal solution.
Section 4.2: I can understand that this is super-interesting and it was possibly fun to write all these things down. But here, I think, the paper goes beyond what it is meant for - a review of existing data sets and processing approaches. As I detailed in my very first comment, 1-2 key studies per observational data set, emphasizing its relevance and its application potential are very fine - but placed upfront in your study.

I note that a lot of what is written herein is actually a repetition of what you have written already?
L703-707: Isn't this statement downgrading the importance of the SHEBA and MOSAiC drift campaigns? It believe the role of these expeditions - and potentially of other expeditions of shorter duration but with investigations on drift ice - needs to be emphasized better.
L757-762: While what you write here is true ... there is little hope to better and reliably disentangle the knot caused by the similarities in SAR signatures between melt-pond free sea ice, melt ponds, and leads - at least to my understanding.
L765: Here I would like to make two comments. The first one is that, yes, sea ice extent (still) is the main metric that is used to describe the temporal development of the sea ice cover. But physically using sea ice extent is the poor men's Approach to do so because the main, physically correct quantity is the sea ice area. One could therefore ask how relevant melt Ponds are for sea ice area predictions - unless nobody dares to predict that quantity. The second comment is: When it comes to using observational data, then melt ponds don't influence computation of the SIE because the SIE is computed from grid cells with SIC > 15% NO MATTER what the SIC actually is. Hence, SIE is largely independent of any bias in SIC due to melt ponds (or other melt-induced changes in the surface properties of the sea ice). What is true is the SEA ICE AREA (SIA) which takes into account the actual SIC values is influenced by melt ponds. And I believe this lack in SIC accuracy has made the community to divert from SIA to SIE - even though SIE has been shown to not be a credible quantity in this context.
L863-865: Three comments here. First: Are images from these commercial satellites openly accessible? Most of them no. How much of the Arctic sea ice do they cover?

Secondly: I think there is an impressive amount of such data available. A paper like this is very timely to inform about these data sources and I suggest to emphasize this in this paragraph.

Third: You should not forget that any digital image, photography or other instrumentation that is used from aboard a ship or an air-borne device or even the two mentioned satellites requires a retrieval. All these sources are NOT in-situ measurements such as a sea ice thickness measured with a ruler stick is when it comes to evaluating sea ice thickness satellite observations. Therefore one line of action, in my eyes, clearly also has to be to communicate uncertainties for all these "reference" data sets. It needs to become clear that using melt-pond fraction derived from air-borne data such as OIB to evaluate a satellite-based retrieval has the nature of an inter-comparison of two derived or retrieved products and not the nature of validating a satellite product with a set of in-situ measurements. The limitations of use need to be stated clearly.
L885-890: While this is all true, I don't think that the effort that is potentially required here is worth the outcome. This is certainly a very subjective opinion. But as far as I can see it it will always be good to have two or even three such data sets for intercomparisons at hand to also have a means of the uncertainty.
Appendix B: Some of the names here need to be checked. TransArc is for instance a data set that is possibly also included in what you call IceWatch which is a collection of visual ship-based observations. I invite you to check whether IceWatch data are really available until present.
PANGEA, the Niehaus et al data set should be named differently because a number of the data sets you are listing here are available via the PANGAEA data base.
You might want to check this link as well - it is a collection of ship-based observations that starts before 2006 (compared to IceWatch): https://www.cen.uni-hamburg.de/en/icdc/data/cryosphere/seaiceparameter-shipobs.html
I am not sure whether this list is supposed to be exhaustive. I note that in Appendix-C there is another list of data sets and I am a bit puzzled why you do have these two Appendices. Would it make sense to combine them? If you remove the "open source" part in the heading the content should be able to be displayed in a common appendix.

Citation: https://doi.org/10.5194/egusphere-2025-4480-CC1

Sara Aparício, Simon Driscoll, and Daniela Flocco

Viewed

Total article views: 454 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
339	98	17	454	10	9

HTML: 339
PDF: 98
XML: 17
Total: 454
BibTeX: 10
EndNote: 9

Views and downloads (calculated since 09 Oct 2025)

Month	HTML	PDF	XML	Total
Oct 2025	254	45	14	313
Nov 2025	85	53	3	141

Cumulative views and downloads (calculated since 09 Oct 2025)

Month	HTML	PDF	XML	Total
Oct 2025	254	45	14	313
Nov 2025	85	53	3	141

Viewed (geographical distribution)

Total article views: 429 (including HTML, PDF, and XML) Thereof 429 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 21 Nov 2025

Short summary

This study examines how scientists monitor Arctic melt ponds, pools of water forming on sea ice during summer, which absorb more sunlight than ice, creating a feedback loop affecting global climate. It reviews current satellites and field campaigns to acquire data; algorithms used to process it. It analyses over forty studies, comparing datasets, regions studied, and major limitations allowing to identify future research areas to improve Earth observations of melt ponds.


Total:	0
HTML:	0
PDF:	0
XML:	0