|The manuscript describes firn evolution at a high elevation site on Kaskawulsh Glacier, St. Elias Mountains, Yukon. This is a highly relevant topic and the presented results contribute to an improved understanding of ongoing trends in firn density, temperature and potential development of firn aquifers in mountainous environments. Generally, I find that the observational datasets are well described and interpreted. Also, the comparison with model output is valuable. Still, I have some moderate to major concerns that I would like to see addressed, primarily related to the lack of model calibration and the interpretation of surface lowering. If it would be an option to perform some additional model experiments, I would highly recommend to perform some additional runs to calibrate melt rates e.g. by minimizing the misfit between modelled and observed subsurface temperatures. Right now, discrepancies between the model and observations are hardly discussed and the lack of model calibration makes it difficult to draw strong conclusions on trends in firn conditions. My specific comments are listed below.|
L39-41: This needs to be reformulated. The phrase "If the surface continues to melt" should be removed, since refreezing will happen directly when melt water enters cold snow/firn. 'Warming firn' does not necessarily lead to more refreezing, rather the opposite.
L42: A firn aquifer only forms if the water does not directly run off through moulins/crevasses.
L42-44: It would be good to consistently use “perennial firn aquifer” rather than “firn aquifer”, since here long-term (multi-annual) storage of water in firn is meant.
L51-52: Another useful reference for the Canadian Arctic is Noël et al. (2018; https://doi.org/10.1029/2017JF004304), and for Svalbard references to Van Pelt et al. (2019; https://doi.org/10.5194/tc-13-2259-2019) and Noël et al. (2020; https://doi.org/10.1038/s41467-020-18356-1) could be relevant to add.
L55: Here Machguth et al. (2016; https://doi.org/10.1038/nclimate2899) could be cited.
L69: I suppose this refers to air temperature? Please clarify.
L107-117: This part fits better at the start of the next section (3.2).
L137: I suppose L_unc should be dL (or dL in the equation should be L_unc).
L186-190: “Surface lowering associated with refreezing” is confusing. Surface melting leads to thinning and gravitational settling of the snow as well, but refreezing just adds mass to the existing vertical column and does not (directly) cause any thinning. See also my later comment.
Section 3.4: It appears that no calibration of the model has been done, presumably because there were no melt observations to compare to (?). This currently makes it very hard to trust the model output, especially since there appear to be major biases in modelled subsurface temperatures, which may indicate an underestimation of melt rates. See also my later comment.
Section 3.4: Most likely the subsurface model also simulates density evolution, but this is not mentioned here and no graphs of it are shown in the results. It would be an important validation of the model results if it could be shown that simulated density evolution matches the observed densities well.
Section 3.4: I am missing a description of how the model and in particular the subsurface conditions were initialized, i.e. if some spin up has been done.
L249: “heat advection from melt water percolation” seems odd, since melt water typically is at 0 degrees C and if it encounters cold snow it will just refreeze thereby releasing heat. This is the only way in which heat is “advected”, but that is probably not meant.
L251-252: The symbols k_h and k_w are mixed.
L284-293: It remains unclear how melt-affected and not melt-affected firn are distinguished. I think this is an important aspect, because if there is indeed non melt-affected firn with a firn aquifer below, that probably implies that fast deep melt water percolation through piping is an important process here. I would like to see additional discussion on this in the manuscript.
L306-313: What is the density difference between 1964 and 2018 when considering the mean density between the LSS and 15m depth?
L318-319: This is comparing snapshots of subsurface temperature to extract trends. Since subsurface temperatures may vary strongly from year to year, care should be taken to determine long-term trends based on only two snapshots in time.
L344-345: These kind of statements about melt trends are hard to defend without any calibration or validation of melt estimates against observations.
L375-376: “due to increased presence of ice layers”: Is there any information on ice content in the 1964 core?
L381-383: The modelled subsurface temperature trends may be reasonable, but the absolute values are quite a bit off when comparing temperatures in Fig. 5 and Fig. 7. This discrepancy is important to discuss in much more detail in the manuscript, especially since many of the conclusions for example on when firn became temperate and when the PFA may have formed are based on the modelled temperature evolution.
L382-383: “The ERA5 climate analysis”: This is rather an analysis of snow model output. Please reformulate.
L388: “increases” and “effect”
L391-394: I would suggest to reformulate this. It is unclear what this "first stage of densification" is. In my view there are two processes that affect densification 1) gravitational settling (which will go faster at higher temperatures) and 2) refreezing. Refreezing will increase subsurface temperatures, which in turn may increase the densification rate by gravitational settling/packing. That is a completely different sequence of processes than described in L391-394.
L411-412: “The firn model predicted the development of wet,temperate conditions in the deep firn following the 2013 melt season, although it took two years to fully develop (Figure 7).” But the observations reveal that the firn was already temperate in 2006. This should be acknowledged.
L440: “Kuipers Munneke et al. (2014)”
L445-446: Kuipers Munneke et al. (2014) indicate what accumulation and melt conditions favour the development of firn aquifers. So in addition to the accumulation comparison it would be good to also compare melt rates with rates observed in southeast Greenland.
L453: Temperature of the firn will not have a major impact on the perennial firn aquifer. Typically once a perennial firn aquifer has formed the firn above it is temperate and the winter cold wave does not penetrate deep enough to case any refreezing. A factor that is important though is how easily the water can runoff via moulins and crevasses.
L454-455: Internal accumulation commonly refers to the amount of refreezing below the last summer surface, which is probably not what is meant here. See for example Cogley et al. (2011; https://wgms.ch/downloads/Cogley_etal_2011.pdf).
L469-470: Ice layers in snow and firn happen in any accumulation zone that experiences some melt, which is the case for the vast majority of glaciers on Earth. Hence, the presence of ice layers in firn is not something special. Please reformulate.
L474-475: This is an important notion. I understand that there may not be any melt observations to make use of, but I would instead strongly suggest to perform new modelling experiments where one or more parameters affecting the modelled melt rates are calibrated such that a best match between modelled and observed subsurface temperature is achieved. Right now, it seems that modelled melt rates are underestimated, which would result in too little water percolation and refreezing in snow and firn, thereby explaining the currently underestimated subsurface temperatures. With a calibrated model, confidence in modelled melt rates and firn conditions would considerably increase!
L477: “with most of the meltwater refreezing”: It is unclear if this is still the case. The subsurface temperatures reveal that firn was already temperate in 2006 implying that already then some melt water did not refreeze.
L484: It would be nice to have an additional figure showing the modelled density evolution.
L487: "0.73+/-0.23 m". If this is calculated from Eq 6 then, if I am correct, this is not the actual surface lowering, but rather the surface lowering relative to a snow/firn pack that would not experience any melting. I do not really see why this is relevant here. For me, the interesting thing to know would be how much additional refrozen mass sits in the firn column in 2018 compared to 1964, because that is a mass term that is missed by geodetic mass balance observations.
L487-488: “to have experienced a minimum of 0.73 ±0.23 m of surface lowering due to internal refreezing”: Refreezing does not lower the surface, melting does and gravitational settling of snow/firn. Please clarify and rephrase.
L487-497: I am missing a bit the point here. Surface elevation changes are the effect of long-term trends in melt and accumulation. How much of the melt water refreezes does not (directly) affect surface elevation or thinning. I would rather expect a discussion here on the impact of increased densification on geodetic mass balance estimates. Geodetic mass balance observations will just consider surface height changes and not any mass changes that result from an increasing density of firn.
L496: “liquid water retention processes cause the surface to lower”. This is not correct. Refreezing just adds mass to the existing firn column, which leads to densification, but not to thinning! Higher firn temperatures after refreezing do speed up the compaction (gravitational settling) process though.
Figure 1: It could be good to include coordinate axes.
Figure 6c: In addition to Fig. 7 also Fig 6c confirms that the modelled subsurface temperatures are much colder than observed (Fig. 5).