To the editor and authors of “The vertical structure of the troposphere and its connection to the surface mass balance of Flade Isblink in Northeast Greenland”

The authors in this manuscript report the results of 130 soundings by UAV at the Villum Research Station located in Northern Greenland. Their objective is to connect model realizations of surface mass balance to vertical atmospheric profiles of temperature as well as to evaluate the ability of a reanalysis product to represent those temperatures across several “surface types”. In the current state of this manuscript, I believe the authors are unable to accomplish this objective cleanly and I am rejecting the manuscript. Additionally, the writing in this manuscript is poor, carries with it a lack in storytelling, riddled with typos, and isn’t yet to the quality of scientific publication. I recognize that this is the first manuscript of an early career scientist so I wish to get across that they shouldn’t be entirely discouraged. There is a good paper in this work that is worth writing. I encourage the first author to bring this work back to the drawing board and the end product will be something they will be proud of. The below list is non-exhaustive and does not include any technical corrections, but I have compiled some of the red flags which stuck out to me about the manuscript:

The introduction of the text leads me towards an expectation that this manuscript is going to comment specifically on glacier ice sheet mass balance. I agree with the authors on the importance of understanding surface mass balance (SMB) in this context. They likewise mention SMB in other parts of the manuscript, characterizing the ice loss in larger regions of the ice sheet. Why then is 3/4^th of the analysis on non-glaciated parts of the VRS? The framing of the manuscript needs to be rewritten, with emphasis not on a major uncertainty of the Greenlandic Ice Sheet, but on the background needed for what the authors scope of work can comment on.

One major pillar of this manuscript if the comparison of UAV temperature profiles is to the CARRA reanalysis product. CARRA assimilates weather station data as part of its reanalysis product. The authors fail to report that VRS is itself a weather station included in the data assimilation (Figure 2.2.7.1, https://confluence.ecmwf.int/display/CKB/Copernicus+Arctic+Regional+Reanalysis+(CARRA):+Full+system+documentation ). Now data assimilations are never one-to-one matches, but there is significantly less value in evaluating the utility of a reanalysis at the location of its tie points. If a reanalysis is accurate, it is most accurate at the location of a tie point. Still, VRS is only a ground station tie point and thus the vertical profile might still be worth addressing. Regardless, this needs to be acknowledged or used as a guide to direct additional analysis.

The authors have a large operational filter on their dataset that is acknowledged but then disregarded. They are limited by precipitation and wind speed. They may also be limited by time of day (i.e. waking hours) but that isn’t reported. They likewise do not distinguish by clouds, either low clouds which may disrupt their sampling or by higher clouds, which like low clouds would profoundly affect the surface energy balance at the time of sounding. Despite this limitation, the authors claim on Line 311 that “CARRA represents the vertical atmospheric structure around VRS well”. This is entirely inaccurate and not sufficiently reductive to the evidence the authors have to make such a statement. A correct statement might be something such as “Our observation-reanalysis comparison shows that CARRA is accurate in temperature (MAD = XXXX) within  XXXX-XXXX m AGL during 00:00 – 00:00 on clear sky days.” Anything less reductive cannot be demonstrated with the supplied analysis.

The authors utilize an iMet-XQ2 sensor onboard a multirotor UAV to profile the atmosphere. Having used the same combination myself, I know this is an apt choice. That said, what is missing such that I am baffled it isn’t included in the analysis, is the humidity measurement that comes along with the temperature measurement. Humidity is as key of an atmospheric state variable as temperature and likewise just as important to understanding the energy balance of the surface and near-atmosphere such that it is inappropriate to be excluded from the analysis.

The analysis does not include a metrological discussion and interpretation of atmospheric soundings. Discussion on, for example, the location of a surface layer, is missing from the analysis. This omission shines through when the authors arbitrary choose 100m as the lower limit of data excluded from clustering. How was that altitude determined? In Figure 2 (a) the authors show selected average profiles (related: averaged and selected how exactly?) that do not relax to the typical lapse rate until about 200 m for plotted profiles. Why use 100m? The remedy for this is a careful appraisal of atmospheric structure that is part of the reported text.

The authors utilize the MAR model to produce the SMB for VRS. This is an appropriate choice. That said, MAR also includes atmospheric temperature. Why not also include a MAR comparison? The reason for doing so is clear. The authors note that wind direction-based anomalies in CARRA are due to upwind surface impacts and that such effects “may not be resolved in CARRA”. First, I’d like to note that the “may” here can definitely be resolved, as CARRA has extensive documentation. I encourage the authors to spend more time investigating why CARRA could have such a mismatch at a more technical level. Regardless, if the authors chose to use MAR for the quality of its surface mass balance, then it is also the tool to test the impact of easterly winds on vertical profiles.

The authors use K-means clustering to group atmospheric observations. This is an accepted use. However, they fail to mention with what? K-means clustering is typically used in situations which highly multivariable data, though the authors only present data for temperature. My guess is that they do so on CARRA data, as they later define clusters by regional pressures. Either way, this needs to be explicitly discussed in the text before clusters can be evaluated.

The authors concluded that surface albedo affects the surface mass balance at their sampling site. Albedo isn’t a new result and I would hope that the authors would develop a more quantitative description of albedo at their site. Also, given the weight of the importance placed on the concept of snow-albedo feedback, I would expect the word “albedo” to show up earlier in the background or methods rather than for the first time in the discussion on Line 335.

The authors have a tendency to overuse verbiage with value judgements included. Example such as Line 29 “unequivocally” or Line 61 “exceptionally” are unnecessary and not appropriately reductive to the presented and supported science in the manuscript.

The authors tend to report relationships without quantifying them. Examples such as Line 62: What height is near surface, when is the coldest month. Line 112: What improvements? Line 129: How well?

Summary

In this study, Fipper and coauthors analyze the vertical temperature structure above Flade Isblink Ice Cap (FIIC) in northeast Greenland during summer and its influence on the ice cap's surface mass balance (SMB). They first use a large number of atmospheric profiles measured by uncrewed aerial vehicles (UAVs) over difference surface types near Villum Research Station to relate the near-surface thermal gradients to surface type and assess the accuracy of the Copernicus Arctic Regional Reanalysis (CARRA) in reproducing these temperature structures. Finding a good agreement between CARRA and the UAV profiles, they use CARRA to extend the analysis in time and assess the synoptic circulation patterns that control the vertical temperature structure over FIIC. They find that the decreasing SMB trend for FIIC is related to summer atmospheric warming across most circulation patterns, and that the anticyclonic patterns that favor increased surface melt have shown an increasing trend over time.

In my assessment, this is an excellent study overall. The paper is well-written and easy to follow, with clear figures that illustrate the main findings. The use of UAV data to validate the CARRA reanalysis is innovative and the methods for this comparison appear robust. I have some comments below asking for some minor clarifications and technical corrections. Once these comments are addressed, I feel this paper will be a valuable addition to the literature on the coupling of the atmosphere and cryosphere in Greenland.

Minor comments

- General comment: Although no UAV measurements were taken at the top of the FIIC, it might be helpful to have some discussion of how applicable these results are to the ice cap in this topographically complex region. Do the authors expect that the vertical temperature profiles and gradients over the higher elevations of the ice cap would be similar to those observed at lower elevations? And in using the CARRA data for the long-term analysis (e.g. L125–127), are the CARRA data only from the three fixed grid points, or are data from a larger spatial domain around FIIC examined?

- L24–25: There appears to be a discrepancy between the abstract and the findings presented in Section 3.4. The abstract states that FIIC mass loss is driven by "rising summer air temperatures for all synoptic conditions", while L286–287 states that cluster 1 (low pressure) does not exhibit a significant linear trend in either summer mean SMB or summer air temperature.

- L75 (Fig. 1): The Fig. 1 caption states that base layers are from Google satellite imagery – is the imagery actually from a data source like Landsat or Sentinel that is indexed by Google?

- L100: States that the temperature measurements are averaged into 12m elevation bins to account for the Imet-XQ2 sensor's vertical accuracy. Is it possible to get more accurate information on the UAV's vertical position from the UAV itself (i.e. from an onboard GPS or some other data source)?

- L131–134: Similar to my general comment above, is the K-means clustering applied only to the temperature profile data from the three fixed grid points from CARRA? Or to a more spatially extensive dataset?

- ~L170 (Fig. 2): I suggest making the delta-T vertical grid line at 0 thicker, for easier visibility on the figure

- ~L195 (Fig. 3): This figure is really neat. I assume this figure is based on the CARRA data, and UAV data does not enter into it at all?

- ~L197 and elsewhere: Are there no observations from Villum Research Station (VRS) that can be used to quantify snow cover, instead of using CARRA for snow cover? Do the studies cited in L325–327 about snow cover trends at VRS use snow cover observations from the station?

- ~L215: It would be nice to have some more discussion about what causes the varying vertical temperature gradients and temperature profiles across clusters, as shown in Fig. 4. Unless I missed it, there isn't really any physical explanation provided for why these distinct shapes of the temperature profile occur for each cluster

- ~L237 (Fig. 5): I suggest the authors consider plotting wind vectors on the map panels of Fig. 5 (Fig. 5a), to more directly show the synoptic flow direction. For example, L364 discusses northerly wind advection during the cluster 1 regime that could be shown more directly with composite wind vectors.

- L245: How is SMB per unit area different from summing SMB across all grid cells on the ice cap?

- L245: How is SMB per unit area different from summing SMB across all grid cells on the ice cap?

- L304–305: Could there also be a downsloping effect from FIIC that influences the disagreement between CARRA and UAV profiles for the easterly wind regime?

- L309–311: Also, it appears that the the overwhelming majority of the field days were categorized into Cluster 1 (low pressure) or Cluster 2 (zonal) synoptic flow, with little sampling of the anticyclonic conditions? (see Fig. 5d)

Technical corrections

- L250: This reference to Fig. 6c should instead be to Fig. 6b?

- L273: "rare" --> "rarely"

- L349–350: Something about the grammar is off in this sentence... should this be written instead as "...pronounced as *a* disproportional drop..."?

- L380: "Why not also CL1 shows" --> "why CL1 does not also show"