General Comments (overall quality of the preprint)

The subject matter of this paper is highly relevant and worthy of investigation as it concerns understanding the future mass balance of glaciers, ice caps and ice sheets globally. The datasets used in this study, which include a combination of valuable field data (GPR and dGPS), and state-of-the-art firn densification, and flow models should be sufficient to address the main objectives of this study. Scientific rigor however is absent in several areas, particularly in supporting primary conclusions of this paper. For example, one of the main results, ie. PFA has persisted since the 1950’s, is based on reanalysis data, for which there has been no attempt to establish quantitative relevancy to the study site. There are several instances where correlations between datasets, which are not present, would be helpful to provide more informed statements about the conclusions of this paper. Also, there has been very little attempt to explore other evidence to support the primary conclusions, as discussed below.

This paper is fraught with poor writing quality, numerous errors in punctuation, inconsistent formatting, and very poor sentence structure throughout most of the paper. Some sections however are better than others. Also, there are many problems with the figures (discussed below) which make interpretation of the results of this paper challenging. As a result, a thorough assessment of the scientific validity of this paper is not possible.

Specific Comments (individual scientific questions/issues)

L18 (abstract): The statement “ We find that the aquifer was present in 1957, and…”  is unsubstantiated based on the evidence provided. Specifically, the main point of evidence given within the paper for the PFA to be present in 1957, ie. “Given that temperatures during 1957-1977 were likely cooler than during 1937-1957 (Nordli et al., 2020)…”  is not even a certainty so how can this ‘finding’ that the PFA existed in the 1950’s be stated with absolute certainty, as it is in the abstract? Further more, the temperature data used in the model, ie. Riestad, (2011), is downscaled to ocean surface. At the very least, a correlation between the longer term Riestad data (downscaled over ocean surfaces) and the data collected from the Upsalla University AWS (~1000 m a.s.l.), should be stated.

Also, why is Nordii used as the reference for existence of the PFA, where the Riestad data was used to drive the model?

Beyond the model results, this paper does not explore any further evidence to support, or question existence the PFA ‘existed’ in the 1950’s. Is there no other evidence to support, or refute the possibility that preferential water flow routing may have been affected by changes, in ice dynamics, topography, thickness, and/or firn density over this 60+ year period of time? At the very least, it would be helpful to include a figure of near-surface air temperatures from the nearest station in order to provide the reader with some sense of the magnitude of temperature variability over the 60+ year period of this study. Simple calculations of the depth of penetration of the cold wave, and the modeled temperature regime that existed at depths to the bottom of the PFA.

L152 “Furthermore, fast deep percolation is modelled using the parameterization by Marchenko et al. (2017)”. Question: Is percolation of meltwater not being modelled in 2 ways then, ie. by the Van Pelt EBFM AND the Marchenko model ?

L296-298:

 “At the end of our spin-up routine in 1977 a PFA is present in the firn. Given that temperatures during 1957-1977 were likely cooler than during 1937-1957 (Nordli et al., 2020), we argue it is highly likely that a PFA was already present in 1957 and in the preceding decades.”

• This argument that a PFA was already present in 1957, is not convincing as it is poorly supported with scientific proof, or even a convincing argument.

L120: this statement … ‘Firn density and meltwater input to the aquifer is required to model a PFA.’ comes across as a general statement about modeling a PFA. I’m sure this is not what is intended by the author as it ignores many other factors that could be relevant. Please clarify.

L255-260:  re: 2017 non-coherent data…. Could this not have been simulated by deriving the correlation values for the areas in common over the 3 time periods, ie remove all points from  areas in 2017 that do not exist in the 2018 and 2019 datasets. This may provide some quantitative basis to refute or support the reason for coherence in the 2017 data.

Also, could there have been other factors responsible for the low coherence in the 2017 data, eg, towing speed of radar?, rough surface topography?

Technical Corrections

Fig 1: -  Significant place names should be added to the left hand side figure. Include label location of the Svalbard airport (as it is mentioned in the text). The figure caption is poorly worded and hard to understand.

• I don’t see 2015 or 2016 on the figure, only in the caption
• 2017 looks green on the fig, indicated as yellow in the caption
• In sentence 2, its not clear what ‘red rectangle’ you are referring to – left or right hand figure.

For this sentence….“The rectangles correspond to the minimum and maximum coordinates of the measurement in those years and thus show the extend of the PFA measurements” .

• ‘Extend’ should be ‘extent’ (2 instances), turned should be converted,

• Avoid over-use of first-person pronouns - it is very distracting to the readability of the paper. This is particularly an issue in the results section 4.1 and conclusion.

• Please correct citations to use semi-colon, not comma to separate a list of references. An example of this is on line 50, but it happens throughout.

• Please follow the rules of the journal when referring to figures and tables, noting the differences between use of these words at the beginning of the sentence versus in running text. Many errors of this kind throughout the paper.

• Please note that the word "Table" is never abbreviated and should be capitalized when followed by a number (e.g. Table 4).
• You have ‘in situ’ and ‘in-situ’ throughout, please be consistent with formatting rules.

L165: remove the word ‘in’ from the bracketed text.

L215: ‘transfer’ doesn’t seem like the right word… ‘transform’ or ‘convert’ perhaps?

               -remove space before comma

L238:   Improper use of brackets

L216: change ‘will’ to ‘is’

• It is not common for a DEM to have a variable cell size - please clarify.

L218: the differences here needs to be stated. please indicate rmsd between the datasets

L235: close off  should be close-off

L262: no hyphen in northwestern or southwestern

L263: include the max and min differences between the modelled and observed water table depths.

L310; this sentence makes no sense on its own.

L325: should be Fig. not Fig

L337: “If refreezing happens, an ice lens might form.” This is a weird statement -  Why would ice not form if refreezing ‘happens’?

L349: improve the results by how much? Enough to explain the absence of crevasses as being the reason why coherence was so much lower?

L220-240: this section is misplaced as it describes methods, not results.

L252:  Change…

‘The modelled water table is overestimated in the south-eastern corner’

 to

‘The modelled water table depth is overestimated in the south-eastern corner’.

L228: consequence is misspelt.

L320 and beyond, please refer to  ‘density’ as ‘firn density’

Figure 2: y-axis – specify ‘above mean sea-level’

Figure 4. it would be far easier to compare the spatial pattern of differences between observed and modelled if the same color scheme for water table depth was used for both. Or, a third plot could be added to illustrate the differences.

• It would be very useful here to have a box superimposed on the 2017 results which indicates the extent/positioning of the 2018/19 results.
• Caption refers to ‘left column’ and ‘right’.

Figure 5: units missing on both x and y axis.

Figure 7:

• Fix caption (ie., Fig 7 to Figure 7.)
• Color code left hand axis title to match blue line.
• Refer to water table line as blue.
• Not sure you need to specify both ‘right’ and ‘orange’

see uploaded pdf for comments