Overarching Remark: Contribution

I am surprised by such a concise author list for a data paper of this magnitude. While I appreciate that the authors describe assembling and digitizing existing online resources, the net effect of compiling such a data paper can result in citation siphoning from the underlying papers. Simply put, the tendency to cite this data paper will likely reduce the citations, and thus the visibility, of original works. This can be especially detrimental for early-career researchers and/or less well-known research groups. (e.g. https://www.nature.com/nature-index/news/review-articles-cause-dramatic-loss-in-citations-for-original-research) 

When we assembled the Løkkegaard et al. (2023) review paper and associated Mankoff et al. (2022) database, we made a deliberate effort to avoid citation siphoning by following CREDIT principles (Brand et al., 2015; https://doi.org/10.1087/20150211) and introducing explicit guidance in the Data Statement to ensure contributor credit. I gently point out that Jacquemart and Welty (2024) contribute to citation siphoning by simply citing the “Mankoff (2022)” Github working directory, rather than the recommended “Mankoff et al., 2022” GEUS Dataverse data citation that includes the >20 data  authors (see: https://doi.org/10.22008/FK2/3BVF9V).

The current authorship structure of both the paper and the Zenodo repository (both cited as “Jacquemart and Welty (2024)”) makes no attempt to combat citation siphoning. Several times, the authors acknowledge the laborious nature of collecting ice temperature profiles as well as the hopes that glenglat will be adopted as a community database receiving new submissions, but the present authorship structure effectively asks researchers to provide their data without a clear path for receiving citation. Being a “data contributor” on Zenodo generates no formal citation or impact metric. 

Smaller Comments:

L4/L53 -- Explicitly define “glenglat” acronym on first appearance in Abstract and body (Introduction).  

L94 -- It seems a little inconsistent to “occasionally include” snow temperatures simply if they are alongside ice temperatures. 

L117 -- Would perhaps be helpful to have an overview figure of the glenglat architecture, data flow, through different software packages and file types.

L135 -- Is the metadata being described here the specific tags shown in Table 2? This can be made explicit. 

L164 -- It would seem appropriate to have description of the digitization error uncertainty (i.e. Figure 10) here, as it is part of the Method and not Results/Discussion of the data base?

L205 -- Yes, it does seem relevant to provide in the database where boreholes are in the accumulation or ablation area. 

L225 -- It would be useful to visualize/assess whether warm bias changes with time, as presumably more recent records have a greater likelihood of meltwater refreezing within firn, and releasing latent heat, than older records.

L282 -- Perhaps describe a couple examples of prime targets for such a retroactive comparison?

L315 -- Some citations for the cold-temperate transition being “often extracted” from radar data.

Table 3/4 -- Please clarify what “/**/” denotes in the file extension. 

Figure 6 -- It may be helpful to indicate which profiles are temperate and cold-based in 6b, as temperate ice is an upper limit temperature. It might also be insightful to divide the comparison into 1950-1990 and post-1990 (or c. 2000 breakpoint) to highlight if the biases are different in the more recent period during which more meltwater is percolating into historical firn accumulation zones than in the past. 

Section 2.4 -- What do you do when the profile graphic to be digitized simply has a line graph, without specific points, and thus the individual measurements are not discretized?

Section 3.2 -- The comparison against ERA5 air temperature and precipitation is interesting, although I would caution that more caveats need to be provided with regards to interpreting the ERA5 data, which is known to be challenged in areas of complex topography -- where most glaciers are often found. For example, could ERA5 bias with elevation contribute to the apparent warm bias? Comparing modelled surface temperatures with observed ice temperatures is different from comparing observed surface temperatures with observed ice temperatures. Also, be explicit on the temporal subset of profiles in this comparison, which appears to be only post-1950 profiles.

Section 3.3 -- I am not sure this involved discussion of sampling bias adds much to the paper. Yes, a sampling bias clearly exists, it is not clear why it is important. Would the authors perhaps expand on how they envision -- at highest level -- the database potentially being used in such a way as these biases become important? For example, if used as a training dataset when simulating the global population of englacial ice temperatures, would the bias result in potentially over- or under-estimating either present day or historical mean ice temperatures?

Section 3.4 -- There is a very nice discussion of the digitization error here. Perhaps readers would appreciate more discussion (or guidance) on the population of measurement errors reported in Figure 10, and whether some generalization can be made for an adopted value. Or if the mention of 0.14°C as the mean meant to reflect a suitable “characteristic value” for the measurement uncertainty population? It would also be helpful to have further discussion on how different types of errors interact or combine to potentially bias results, and how to mitigate this issue.

Section 3.4 -- It would also be helpful for some discussion of depth uncertainty. The database is effectively T(z), and there is clearly uncertainty in z that is not captured at present. There is both digitization uncertainty and also measurement uncertainty, especially when boreholes do not reach the bed and have poorly constrained altitudes. Presumably, end users will need guidance on how to compare a historical depth temperature with modeled glacier geometry. For example, when writing Løkkegaard et al. (2023), we were urged to also express temperatures on fraction 0 to 1 depth scale that could be fit to modelled ice geometries. 

Data License -- I understand that the CC Attribution 4.0 International license ensures, among other things, “appropriate credit” to the current work (i.e. Jacquemart and Welty (2024)) and transparency for any changes to derivative products (https://creativecommons.org/licenses/by/4.0/deed.en). Authorship aside, I am not entirely sure if glenglat itself has met these data conditions with respect to the original works, particularly when digitization of a figure has created a derivative dataset of an original profile for which no DOI is available.  Is it possible to make otherwise unavailable .pdf's available in a protected literature repository? This is an approach I have seen in other community data compilations, which can also help elevate otherwise overlooked works. That could address the appropriate credit link to the original work. For transparency on the derivative product, perhaps showing the original figures as well as the original final with the digitized profile overlaid? This would highlight how well the dertivative fits the original.