Review of Steiner et al

The terrestrial ice margin morphology in Kalaallit Nunaat (Greenland)

The authors state two goals. First, to provide the lengths of margins of the three main categories (marine, land, and lake terminating).  And a second goal to show the variable morphology of the land-terminating glaciers, which range of shallow ramps to steep cliffs.

First - I really like that this paper is tackling the question of land-terminating glacier terminus morphologies. There is lots to be learned from the terminus behavior. So I really appreciate the authors efforts!

This paper overall describes the data set, and does not provide any scientific analysis alongside that data set. I admit that I expected - based on the title - to see some interpretation of the data. The last sentence of the abstract informs reader that this paper describes the data set and they leave the interpretation to others.

What I see as unique about this paper is the focus on the varying morphologies of the land-terminating ice margin. This goal, however, seems to take lesser importance in the abstract to the overall lengths of the margins. I would suggest that the authors emphasize this more novel data set and less on the cumulative distance that is marine/lake/land.

First the general categorization goal: While the categorizing of termini and determining total lengths of each can be important for some projects, without context for the numbers provided in the abstract, this is not so useful. Are those numbers bigger or smaller than expected? Are they changing? What year do they represent? I appreciate that they include all the margins, not just those with significant ice discharge - as the lateral margins of glaciers can play an important role. I looked to one tidewater glacier as an example in the data set - one where I have an upcoming project (69.79521923, -50.23659187) and saw 3.5km section of the northwest lateral margin (presumably against a fjord wall) flipping between land and tidewater, of a tidewater glacier as a mix of land and tidewater termini. Similarly the southeast margin has a section identified as tidewater that is not. This makes me question the overall quality control and the value of publishing these numbers. The authors state in line 123 that they did visual quality control, so I am surprised that one random glacier picked seems to have errors - as does the glacier to the north. The data set is useful! - although specific use might require fixing the errors. For example, ice flux could be more accurately determined for different regions by the dot product of the normal to the margin and the ice flow direction across the margin - if both data sets are of sufficient resolution.

My suggestion is that this primary margin categorization be less of the end goal, and less emphasized in the abstract - unless the authors want to interpret their results in context of the Ryan et al 2023 and other papers that have done similar work using a slightly different approach and goals. And please be more clear that quality control is weak (i.e. line 123) - so the data should be used with caution.   

Towards their goal of the shape of the land terminus - this I find is a unique and useful contribution, as the shape of the terminus reflects the interaction of physical processes (ice flow with fracturing and details of surface mass balance). And I believe we can learn about evolving ice dynamics from these margins.

The paper describes the methods for delineating steep versus shallow ramps in a clear way - I am not an expert in remote sensing analysis of this kind, so I don’t feel like I can speak to those details, but it generally seems robust except for the issue of it being clear what dates at attached to the results, so they can be used to look at changes.

If the authors choose not to do any analysis with their data set, I would suggest one additional statistical result might be helpful for understanding the data - perhaps for quality control, perhaps for science. That is to look at the distribution of lengths of margin segments of each type (separately doing the land/lake/marine versus the steep/shallow). For the general margin categories, this kind of statistical analysis might point to errors - short sections of marine terminating margin <500m that are separated from a longer section of marine terminating might point to places where the categorization is in error (such as the location I identified above). 

For the steep/shallow categorization on the other hand, the distribution of many short segments versus fewer longer segments might offer some insight into the dynamics.   

General comments about the structure:

The introduction spends a lot of time motivating why we want to know the margins - especially the land terminating margins. That approach seems good, but the discussion and conclusions don’t really come back to these ideas - instead the discussion focuses on lengths of different categories and the value of Arctic DEM. The discussion/conclusions that relate back to the introductory material is very general - mostly saying this is future work. My suggestion is better balancing the physical explanations introduced early in the paper with the discussion/conclusions so that there feels like some closure for the reader. If the authors want to emphasize that the biggest gap in answering physical questions is the lack of good data set - then emphasize that more in the introduction, why has there been a lack of sufficiently good data sets to tackle the question of steep versus shallow? Is it just the morphology data set? What other data sets do we need to answer some of these questions? While they suggest this data set is the primary one lacking in line 80, they could expand on this more - and clearly lay out the gap in knowledge they are trying to fill. To summarize my thinking here - the introduction suggests that the authors are going to assess the physical processes more than they do. The discussion/conclusion don’t really say much except that their data set might be helpful for future work (without many details).

I felt like the paper was overall well written and explained their method clearly, I do not have many suggestions details of wording in these areas.

I am happy to chat more with the authors if they wish. 

The Steiner et al team deliver the first Greenland-wide quantification of land-, lake-, and marine-terminating margins and a spatial census of terrestrial margin morphologies for both GrIS and Greenlandic PGICs. This is a major accomplishment, and a very usfull scientific contibution for numerous future process and change detection studies. The paper’s immmediate reproducibility makes it a foundational dataset and method for ice margin research in general.

This manuscript is clear, well-structured, and methodologically careful. The authors combine established datasets with sensible preprocessing, and a distribution-based slope classification, validated against Pléiades DEMs. Uncertainties and limitations are quantified and transparently handled, and all code/data are made available.

I really applaud the authors for doing this work and for taking the time to develop the method. I am confident, that it will be a method and procedure, which will be used on many datasets in the future, and provide critical knowledge and understanding of our ice margins and their changes.

I have one major point, which requires attention:

My major point of concern with the study in its current form is the use of the PROMICE ice mask (Citterio & Ahlstrøm, 2013). This particular outline represents a manually derived ice margin based on aerial photographs collected in the period 1978-1987, and not a year-2000 margin as the authors interpret. This discrepancy results in a potential large glacial retreat between the timing of the margin and the timing of the DEM used for the analysis. The authors do initiate a series of  measures to counter the supposed offset from 2000-2012, but the actual off set in timing is much larger.

My concern is that too many cells have been excluded as a result of this, and only regions where retreat since the mid-1980s have been minimal are included. Under all circumstances, a 100 meter buffer from the 1980s margin, will many regions be inadequate as reported retreat rates are often in the order 5-20 meters / year. This concern is illustrated by figure 4d, where some of the excluded 1km grid cells, show a frontal retreat (between ice mask and DEM) of more than 500 m. Here most of the margin is excluded, and as a result, knowledge of ice margin slope of the rapid retreating ice margins are omitted.

There are a number of ways to go about this, but unfortunately I don’t see any that does not requires substantial extra work. There are newer datasets available that offer an ice margin, closer in time to the ArcticDEM. One option is “OpenLand” from the Danish Climate-data Agency. This outline is from 2017-21 https://dataforsyningen.dk/data/4771. Another option is a beta-dataset from GEUS – the new PROMICE-2022 ice mask, which is currently under review, but with data available https://essd.copernicus.org/preprints/essd-2025-415/, however this is only covering the ice sheet and not the PGICs. This could however be combined with one of the newer PGIC outlines like the Randolph Glacier Inventory which match closer in time to the ArcticDEM.

If the authors argue that the 1980s outline is sufficient, based on the extensive  three step approach with visual inspections, I would like to see a more comprehensive analysis of the effects of the excluded ice margins: How much of the margin is visually inspected? What are the cut-off values for a cell to be discarded? What would be the effect of a different buffer size?

I would expect the excluded cells to be in the lower elevation parts of the terminating margins (the glacier front), which will skew the results more towards lateral margins, with potential substantial implications for the overall results and conclusions.

Many scientists would want such a study to also include change over time, and given the multitude of datasets available, it is indeed also possible, eg using the AeroDEM (Korsgaard et al, 2016) which corresponds exactly in time to the 1980s PROMICE-ice mask. However, I don’t see it as a prerequisite for publishing this study. The work in itself, and the dataset, is sufficient to warren a major scientific contribution, and I am confident that several later change-studies will develop from this paper.