Review of Machguth et al. 2024

This paper provides an update to a methodology to determine runoff limits on the Greenland Ice Sheet from MODIS and in addition investigates differences in runoff from these observations as well as MAR and IMAU-FDM models, partially in RACMO.

While there is clearly some new science in this paper I would argue it needs some major revision if it seeks to answer the question posed in the paper title. The research goal of the paper is unclear, is it to introduce an improved method for using satellite data, or to compare different methods for calculating runoff?  I would argue that for either improvements are needed, due to the following reasoning.

1. The new methodology for improving detection of runoff limits from MODIS is not compared with other remote sensing methods, or with the previous method in Machguth 2022. Although the improvement here is that the method can now be used in more areas, without any validation or comparison it is impossible to judge the validity of the method.
   
   
   
   
2. The authors make suggestions as to why MAR and RACMO/IMAU-FDM may differ but can’t actually evidence this. An RCM and a firn model are very different things, and while it is clear that there is a difference in the implementation of vertical water percolation between the two, that is not the only difference between them. A firn model is run on a very different vertical resolution and will be dependent on RACMO forcing in this case, which will also influence the runoff. The bucket method may only be part of the story as the paper does not compare like for like, or truly compare the two RCMs, or explain why they differ from MODIS completely.

Below I provide line by line comments to add detail to the above.

Title- The paper doesn’t really compare RCMs (plural). It jumps from sometimes including RACMO, sometimes IMAU-FDM forced by RACMO, but to really compare RCMs a full comparison of MAR and RACMO would be needed.

Line 9- The paper does not demonstrate that the difference in implementation of the bucket scheme are responsible for the disparity, only that this is a possible explanation posited by the authors.

Line 18- Suggest addition of ‘among’ our most advanced tools here. Remote sensing, firn models etc. are all advanced tools that contribute to our understanding.

Line 19- Do all these papers actually say RCMs are our most advanced tool? E.g. are they not a part of some of the methodologies described in IMBIE?

Line 46- Confused by the use of the word ‘oppose’ here, how do you oppose the runoff limits?

Line 46- Landsat is mentioned here- why would MODIS be used instead of Landsat? The context is missing here, and a reference for any work done with Landsat as well as justification for why the results in this paper are not compared to any results using Landsat.

Line 75- This is further evidence that MAR vs IMAU-FDM is not a straightforward comparison and that there may be other reasons for differences between them.

Line 80- Again this makes me more confused why Landsat isn’t used, or at least compared against for the MODIS methodology.

Line 94- Does the new method give the same results for the areas covered in Machguth 2022?

Line 105-‘used no more’ or better might be ‘not used’.

Line 105- Is the difference between clean and dirty ice a function of water depth? This doesn’t quite make sense, clean ice, dirty ice and ponded water are different things.

Line 137- Could the differences in albedo scheme also contribute to the differences in runoff found between this model and MAR?

Section 3.2 Please state clearly how IMAU-FDM does (or doesn’t) deal with ice lenses given this is a key difference with MAR.

Line 172- Could the lack of masking for smaller aquifers influence the results?

Line 184- How do the detections in the Tedstone paper compare to those made here?

Section 4.2.1  I found this section and the jumping between RACMO and IMAU-FDM confusing. I would suggest either having a full comparison of MAR, RACMO and IMAU-FDM, or removing the RACMO here as it’s not clear what it adds when it’s only partially included.

Figure 7- This figure to me is the one that really made me question the comparisons made here. It shows very clearly that MAR and IMAU-FDM are working on very different scales and thus capturing different processes, and the suggestion that differences are due to bucket methodology over simplifies this.

Line 311- Is this a lack of inertia or just that processes are not accounted for?

Line 318- This paragraph is confusingly written. It starts by comparing RACMO and IMAU-FDM, then compares MAR and RACMO, but not IMAU-FDM and MAR. Stating that RACMO and IMAU-FDM show similar temporal patterns here is hardly surprising given that one forces the other. The RACMO firn model is also mentioned here but hasn’t been detailed before.

Section 5.2- As stated above I don’t think this paragraph shows the bucket scheme is the main cause of deviations. Several other differences between IMAU-FDM and MAR are mentioned but without justification as to why they are less important.

Line 386- Proof that this is an improved method is missing. It might cover more areas but it is not validated.

Section A2- Please make clearer how this differs from the 2022 paper.

see attached pdf.

The manuscript "Runoff from Greenland's firn area – why do MODIS, RCMs and a firn model disagree?" by Machguth et al. describes a study where the authors developed an improved algorithm to estimate the runoff limit using MODIS. This result is compared to simulation output from firn models in RCMs. A mismatch is found in the extent of the runoff area between MODIS and the modelled runoff area, as well as discrepancies that are present between firn models. This is then investigated further by analyzing detailed firn model output along a transect, which reveals that the way water retention is treated is the main cause for the discrepancies.

The study is very relevant: meltwater runoff from the ice sheets is a major souce of sea level rise, and is in fact quite uncertain (as clear from this study). The study is informative for the further firn model development acitivites in the firn community, and I think it's well suited for publication in The Cryosphere. Nevertheless, revisions are necessary, to improve the substantiation and clarity of the results. I think that a bit more analysis may be required to make the study more relevant. Now, the detailed analysis of possible causes for the model discrepancy is restricted to 1 transect only, which makes it uncertain how well the findings translate to the Greenland ice sheet runoff area as a whole.

Main concerns:

- My biggest concern with the study is that basically only 1 transect is used to substantiated the conclusions. The argument that this transect has been used very frequently in studies, given the wealth of detailed field observations available is a bit weak, because very little field observations are in fact used. For example, no density-depth profile is shown to indicate if the density profile simulated by FDM or by MAR is in closer agreement with field observations. The majority of the results section describes the discrepancies, rather than explain them (as the title "why..." would suggest). RACMO FDM shows a strong transition from basically full ice to firn with substantial pore space around -47.5 longtidude. Immediately, this raises the question if this is general behaviour from FDM, also found further north? I think that the authors should try to find a way to make the discussion and results more robust, by analyzing larger parts of the ice sheet. Similarly, Fig. 2, A1 and A2 basically show one flowline. Are the results robust and can be extrapolated to other flowlines?

- An aspect that I think is slightly difficult to grasp for readers not familiar with the topic, is how the different definitions of runoff are used. When it comes to modelling, a very clear definition is water that leaves the firn column. But here, a crucial difference is that in FDM, water can only leave at the bottom, whereas in MAR, it can leave at the bottom or laterally. Furthermore, bottom in FDM is much deeper down than in MAR. When it comes to MODIS, runoff is estimated from the slush limit, an assumption that is reasonable, but doesn't come without caveats. Maybe a sketch can help to establish clear definitions. Note that in the text, it is is not clearly defined what runoff is for the used firn models. For FDM, it is not mentioned, and for MAR, only lateral runoff is explained. But it should also be explicitly mentioned that water leaving the firn column at the base is considered runoff.

- Fig. 5: Here, I with panels l and k (what happened to panel i and j?) should also include water retained in the firn column as liquid, not only the refrozen part. I think this needs to be included to better explain how the differences in water retention parameterization are responsible for the differences. Moreover, I wonder to what extent it plays a role that the FDM simulated firn column is much deeper. For example, the additional refreezing in FDM shown in 5l is partly caused by the fact that the firn column is deeper in FDM than in MAR. Maybe these figures should be standardized to only show the uppermost 10, or 20m.

- A bit in line with my previous comment: I find it hard to understand how the thermodynamics are different between MAR and FDM. Given that both are driven by ERA5, I assume that the overall energy balance should be more or less similar. The 10m-depth temperature, however, varies largely between MAR and FDM (Fig. A3). I find it very hard to grasp if this is only due to the percolation scheme, or that the firn is generally warmer, or if the surface energy balance is higher in MAR. Also, I think it easily gets confounding that the firn layer in FDM is so much deeper than MAR. This would allow for much more cold content to be stored at depth in FDM than can ever be captured in MAR. I wonder if the authors can give a bit more insight in this. Maybe calculate the uppermost 10m cold content of the firn layer, to see to what extent surface energy balance, and firn temperatures in general differ?

- I would check for consistency in using the terms "saturation" and "water content". For me, saturation is the part of the pore space taken up by water. Thus, 100% saturation means all pore space filled by water. In contrast, liquid water content is most often defined per volume. Thus, 100% saturation in firn with density ~450, would mean a liquid water content of ~50%. So for example, in L111, both water content and saturation are used in the same sentence, and I'm not sure if the 7% refers to saturation, or to liquid water content. In L126, a percentage of 13% is mentioned, but it's not clear if it refers to saturation or liquid water content, since it's only written "irreducible water". Given the numbers, I think 13% is a value for saturation, not liquid water content. Anyway, I would encourage the authors to thoroughly check this throughout the manuscript, because I think now the percentages given are a mix between saturation and liquid water content values.

Minor comments:

- L8: "where meltwater is routed" --> "to route meltwater vertically"

- L25: "found to perform well" is too general. Performs well in terms of? Or on what variables did it perform well? In terms of mass balance, or calculated melt?

- L36: "over which mass loss takes place". I would add "mass loss through runoff", because sublimation and wind erosion are also mass loss terms.

- L39-40: Obviously, the choice of forcing model can have an affect as well. Forcing both RCMs with the same boundaries removes a large part of uncertainty that would come from the GCMs. This is not really a drawback, because it allows to compare RACMO and MAR on more equal footing. But maybe a brief remark on how well ERA reproduces Greenland climate and is suitable to use as forcing along the model boundaries could be justified here.

- L74: I suggest to specify that these concern output time resolution. Like: "MAR output and RACMO 1 km downscaled data" and "Output from RACMO and IMAU-FDM are at ..."

- L134: "150 vs 3000 layers". Maybe specify if this covers equal total firn depth?

- L153: "is rather insensitive": this is actually an interesting point. A figure and a full blown sensitivity study is not necessary, but maybe a statement like: using a threshold of X compared to Y affected the runoff limit only by Z kilometers would be useful here.

- L205: "The plateau is shorter in the north than in the south." is a too general statement given what is shown in the figures. In fact only one transect in the north or in the south is shown. Can this indeed be generalized?

- L171: The way it was written, this actually confused me a bit, because I was looking for the masked points in Fig. 1. I would maybe write it more explicitly, like "Fig. 1 does not show retrievals from 60 to 68 ...."

- L245-246: "and thus could not warm further" is a bit poorly phrased, since this would also limit refreeze once temperatures reach 0 degC.

- L311: Is it really inertia? The example given in the next sentence sounds to me more like non-linear behaviour.

- L347: Please correct: "A secondary reasons"

- L393: "later water flux" --> "lateral water flux" (I assume?)

- Please avoid the use of green and red in one figure for color blinds (Fig. 1 and A4)

- Generally, I think the figure captions are too short.