General comments

The manuscript reports on the temporal evolution of surface velocity of the glacier Vernagtferner in the Eastern Alps with focus on a period of major slowdown in recent years after the glacier broke up in several parts. The presented velocity data are based on in situ stake measurements and manual feature tracking in optical airborne images. The presented material and related discussion cover two main topics: (i) the presentation and description of the derived velocity field and of factors responsible for slow-down, including a review of the glacier behaviour since 1966; (ii) an overview on specific remote sensing products on glacier velocity, leading to the conclusion that these products are not suitable for application to slowly moving glaciers. Topic (i) is an interesting case study on a glacier in retreat, demonstrating the impact of long-term negative mass balance and related glacier thinning on a previously quite active Alpine glacier. Factors responsible for the decline are discussed, but an in depth assessment of the governing processes is missing. The presented material on topic (ii) does not contain any novel aspects. In summary, the paper presents an interesting case study on slow-down of an Alpine glacier in decay but has substantial shortcomings. Major revisions and shortening (in particular regarding topic ii) are required.

Specific Comments

Satellite remote sensing: Major parts of the manuscript refer to spaceborne remote sensing of glacier velocities (Introduction line 45 to 53, Sections 3.2, 3.3.1, 4.3.1, Section 5.2, Fig.2, Fig. 9). These sections refer to properties and suitability of particular products, but do not provide any novel information on methods, accuracy assessment and constraints regarding remote sensing techniques and products for mapping glacier surface motion. During more than three decades results of detailed performance analyses on glacier surface velocity methods and products have been reported in publications and product specification documents, both for so-called “standard satellite remote sensing products” as well as for products derived from data of different satellite missions. These publications show that the pixel size and errors of the “standard products” (based on offset tracking) do not match the accuracy and spatial detail required for mapping very slow velocities at comparatively small spatial scale as observed on Vernagtferner. Taking this into account, the sections on satellite remote sensing can be largely shortened and replaced by references on documents and publications specifying performance numbers for specific satellite-based velocity products. Furthermore, Figures 2 and 9 can be omitted because there is no need showing examples of deficient velocity maps using input data that are not matching the technical requirements needed for velocity retrievals of the study glacier.

Analysis and interpretation of surface velocities: The new velocity data, presented in the manuscript, refer to the period 2018 to 2023 when maximum velocities of Vernagtferner were below 5 m yr^-1. The presented velocities are horizontal displacements referring to individual points on a main branch of the glacier, based on manual feature tracking in optical airborne images and on stake measurements. Considering the average velocity of 1 m yr^-1, it is obvious that the magnitude of vertical surface lowering exceeds that of the horizontal displacements. Consequently, the vertical displacement of the surface (mass depletion due to surface/atmosphere exchange processes) is the dominating component for the glacier mass balance in the current state and ice dynamics plays a minor role. In this context, quantitative information on the annual mass balance (respectively the related topographic change) and its spatial pattern during the study period would be of interest.

Line 45 to 53: This is a one-sided introduction on space-borne remote sensing applications for ice velocity monitoring. The statement on the use of space-borne remote sensing “particularly in Antarctica” does not reflect the actual situation in which ice velocity products are generated routinely on behalf of various organizations, covering at large all global land ice areas. Several of these products exploit also radar repeat-pass interferometry in regions and seasons where coherence is preserved.

Line 56-57: Temporal decorrelation is not a particular problem for slowly flowing glaciers, but rather for fast movement, particularly in shear margins where interferometric fringes are often tightly spaced or aliased.

Line 86: Horizontal displacement and surface velocity are not the same.

Line 113-118: Whereas time spans of TerraSAR-X repeat-pass data are 11 days, there are other high resolution SAR constellations offering shorter repeat-pass sequences, well suitable for glacier velocity mapping based on the motion-related interferometric phase. For example, successful glacier monitoring applications have been reported for 1-day, 3-day and 4-day interferometric repeat pass pairs of the COSMO Sky-Med constellation, providing high accuracy velocity products. Coherence tracking applies cross-correlation matching of templates and thus has a spatial resolution and sensitivity similar to feature tracking.

Figure 3: Please provide information on the time periods (years) to which the time sequences of the individual stakes refer.

Figure 5 caption: Please provide reference for the source of the ice thickness data 1970.

Section 4.2, Seasonal variation: The summer velocities are based on measurements in summer 2022, the ablation period of the mass balance year 2021/22 with the largest mass deficit of the 2018 to 2023 period in the Eastern Alps. Consequently, it is unclear if the number for the seasonal velocity increase 2022 is representative for the whole period.

Figure 8: The presentation of a few numbers (velocities, exact dates) in the form of a table would be more appropriate than the display within a diagram.

Line 200: The statement that it is “not possible to (semi-) automatically produce a reliable surface velocity map from aerial or satellite imagery for slow-flowing Vernagtferner ….” is rather speculative, not taking into account capabilities of advanced airborne and spaceborne observation systems and analysis techniques. For example, several large satellite constellations with very high resolution SAR sensors are in space since several years. Some of these constellations provide repeat interferometric observations of excellent quality with repeat-pass intervals from one day onwards, as for example ICEYE InSAR products show.

Line 223ff: Taking into account that the annual melt losses, amounting up to several meters, may cause significant changes of surface features, the estimates of feature position accuracy seem to be rather optimistic. For example, in line 53 it is stated that “surface features (e.g. crevasses) change considerably during this period”, a possible source for increased uncertainties in feature tracking. Furthermore, oblique views sideways of the central flowline may introduce errors, in particular if the surface elevation at the time of the survey is not exactly known. Please provide information on the procedures in which way these issues are taken into account.

Section 5.1, Ice dynamics: Basic mechanisms related to slow-down of glacier flow are addressed, as well as possible causes for seasonal variations. However, quantitative estimates on the impact and magnitude of the different processes at the study glacier and their interactions during the observation period are missing. For contributing to the advancement of understanding of processes governing the slow-down of retreating glaciers, quantitative estimates would be essential. Furthermore, hints on the significance of the study results in respect to the general glacier behaviour in this region would be of interest.

General comments

The manuscript by Dobler et al. analyzes the ice surface velocity of Vernagtferner, as an example of a well-monitored slow-flowing mountain glacier.

Leveraging 60 years of annual observations of stake positions, the manuscript draws a link between the glacier mass balance history and the patterns of ice speed-up and slowdown. After concluding that existing remote sensing products fail to resolve the slow flow field of Vernagtferner, the manuscript presents a map of present-day surface velocities over 2018-2023, compiled by interpolation of multi-source point measurements such as stakes and manually tracked features.

The topic of remote and in situ ice velocity measurements on slow-flowing glaciers is certainly of current interest, and a report on the long-term dynamics investigations at a data-rich site is definitely relevant and holds significant potential to advance knowledge of the ice dynamics of mountain glaciers.

However, the main direction of the manuscript is not fully clear in its current form - the two covered topics (re-analysis of long-term stake data and compilation of present-day velocity map) are only weakly linked. As such, the main message and achievements of the manuscript are somewhat hard to understand. Moreover, the manuscript stops short of advancing the current knowledge on the topic of ice dynamics. The glaciological conclusions from the analysis of such a rich historical dataset are somewhat qualitative and generic, and in some cases are not adequately supported by the collected evidence. Much of the data needed for interpretation of the ice dynamics (such as historical local mass balance, or changes of ice thickness and surface slope) are not adequately presented in the manuscript. Furthermore, the proposed method to compile a velocity map (in part) from manually tracked point measurements is affected by some significant flaws, raising questions about its suitability and advantages compared to state-of-the-art automated methods. Finally, the uncertainty analysis relies on several arbitrary estimations and assumptions, rather than existing, well-established methods to quantify uncertainty in remotely sensed glacier dynamics.

In light of this, I would suggest to repeat some key parts of the analysis, after reviewing the literature for the most appropriate methods to be applied to the high-quality datasets of Vernagtferner. I would suggest resubmitting the manuscript to reflect these major changes. In particular, I would suggest adjusting the scope to focus more on just one of the two topics - either (i) the compilation of the present-day velocity map, or (ii) the analysis of the historical dataset and its glaciological interpretation. The two topics are quite loosely related and the manuscript would really benefit from a clearer message, answering one or more well-defined research questions.

For topic (i), I would specifically suggest application of digital image correlation on the UAV, aerial and/or satellite imagery, with an appropriate pipeline for pre-filtering, post-processing, and aggregation. Multiple studies (e.g., [1], [2]) have shown that data processing specifically optimized to a study site can resolve the ice dynamics of individual glaciers much better than global products. Given the high-quality available data (such as four years of end-of-season, whole-glacier airborne photogrammetry), I expect this processing to resolve very well the slow movement of Vernagtferner and automatically produce good-quality velocity maps, possibly even at annual intervals and on the whole glacier body, contrary to the current manuscript's conclusion that "It is obviously not possible to (semi-)automatically produce a reliable surface velocity map from aerial or satellite imagery for the slow-flowing Vernagtferner". The stake measurements (and possibly the manually tracked feature) would be a valuable reference for validation and uncertainty estimation. This kind of manuscript might be most suitable for a data-focused journal.

For topic (ii), I would specifically suggest a more thorough re-analysis of the very interesting historical stake data, possibly including: (1) a more realistic model approach than the shallow ice approximation, for example an IGM inversion, given the high-quality data available over the entire glacier; (2) a better-processed ice thickness map, without the obvious major artifacts visible in Fig. 7; and (3) a more detailed analysis of the interplay of surface slope, glacier thickness, and stake velocity anomaly. The recent measurements shown in the manuscript could still be mentioned to compare spatial patterns over time and to investigate seasonality. This kind of manuscript would be most suitable for a glaciological journal.

In both cases, it is important to perform a quantitative, data-driven uncertainty analysis, based on leave-one-out validation and rigorous error propagation, as described in the references provided below.

Specific major comments:

- Presentation of the Data and Methods needs major restructuring in order to support the analysis. In the current form, it is really hard to understand where some important data come from or how they were processed. In particular, it is important to provide full information on: (1) the data sources used for glacier thickness, as well as all processing applied to (i) extrapolate it to the full glacier area, (ii) compute evolution over different time periods, and (iii) extract it at the stakes location; (2) all the stakes (at the very least, all the stakes whose data are plotted in the figures), with their identification numbers, observation period, and maximum/minimum altitude; (3) the actual calculation of modeled velocities according to the shallow ice equation, detailing the estimation and assumptions made for each variable; and (4) the processing of UAV and aerial data, including georeferencing, the used software pipelines, and the resulting spatial resolution and accuracy. Some relevant pieces of information are already mentioned in the manuscript, but in a rather scattered form across the sections, and should be rearranged.

- The discussion lacks almost any comparison of the findings of the present study with those of other investigations, in particular those about (1) long-term trends of ice velocities at reference glaciers in the European Alps, which are monitored at several key sites across the Alpine countries (a single publication on Hintereisferner is currently cited); (2) the derivation of velocity maps of slow-moving glaciers from remote sensing (e.g., [2], [5]); and (3) previous examinations of the dynamics of Vernagtferner (ll. 91-92). It is crucial to situate the findings of the current study in the context of published results, in order to highlight advancements and challenges.

- The manuscript claims that "It is obviously not possible to (semi-)automatically produce a reliable surface velocity map from aerial or satellite imagery for the slow-flowing Vernagtferner", which is the main motivation to propose a manual tracking method (with the resulting downsides for spatial coverage, reproducibility, and labor effectiveness). However, this conclusion is based on observation of some global datasets of glacier velocity, as well as some very poorly detailed testing of automated methods on the available datasets, which are quickly dismissed as an option (ll. 113-119 and 189-199). However, several studies have thoroughly validated the derivation of glacier velocity fields on aerial and UAV imagery with automated methods, such as frequency-domain cross correlation ([3], [4]) and feature tracking ([5], [6]). These methods have been shown to resolve well the ice motion on optical imagery, typically with subpixel accuracy; in particular, the result of [5] (p. 58) clearly shows UAV-based feature tracking fully resolving ice displacements of the same magnitude and time interval as those of Vernagtferner. Thus, a claim of unsuitability of those well-established methods needs to be supported by much better evidence than a quick dismissal, especially given the high-quality available datasets at Vernagtferner (UAV and aerial imagery). The interest of manually pinpointed displacements of individual features is hard to justify without first testing and quantitatively reporting on these state-of-the-art methods.

- The 2018-2023 velocity map is aggregated from data collected over 5 years, on the assumption that ice dynamics would "not change much" over that period. However, Fig. 4 and Fig. 5 show ice velocity changes by more than 50 % taking place over 5 year periods at several stakes. Such a change is larger than (for example) the 30 % correction factor applied to convert summer to annual stake velocities in the data used within the velocity map. As such, the validity of the aggregation of such heterogeneous data is questionable and should at least be discussed in the uncertainty budget. Moreover, the assumption of zero velocity at the glacier edges is questionable - a contribution from transverse stress coupling could potentially be significant on such a slow-flowing glacier. See for example [7]. A rigorous analysis and discussion of these uncertainties is required to support the presented results, especially when the manuscript claims that other existing methods and results are not suitable for the study site.

- The manuscript mentions a "strong sensitivity of velocity to mass balance", claiming that "even the small effect of slightly positive mass balance years on the glacier geometry, results in very pronounced changes of ice flow". While this "effect [...] on the glacier geometry" is not further described, this conclusion suggests that changes of glacier geometry due to positive glacier-wide mass balance would be reflected as faster ice flow in the ablation area already during the same mass balance year, with little to no lag. Such a conclusion is somewhat at odds with the notion of glacier response time and would need to be backed up by some evidence (such as an actual analysis of the mentioned "effect on glacier geometry"). As such, a more thorough analysis of the interplay between glacier mass balance, geometry changes (especially thickness and slope), stake location within the glacier, and stake velocity, is needed.

- Most uncertainty estimations (Sect. 4.4) appear to be qualitative, arbitrary or statistically inaccurate. A more rigorous uncertainty analysis is needed, since an extensive literature exists on relevant methods, specifically concerning glacier dynamics (e.g., [8], [9], [10]). The availability of a large number of data points suggests application of a leave-one-out method for robust uncertainty estimation. Finally, the formulas used to calculate and transform uncertainties should be shown.

Specific minor comments:

- The Introduction needs to provide more focused context on the specific topic of the manuscript, citing more literature on the monitoring of slow-flowing mountain glaciers, on long time series of ice velocities, and on the creation of glacier velocity maps for single glaciers; at present, it is somewhat meandering over broad topics of ice dynamics (basics of glacier flow, glacier hydrology, global ice velocity products).

- l. 29, Nye (1959) was most definitely not the first rigorous investigator of ice velocities; see e.g. [11]

- l. 48, what are "sensors such as Sentinel-2" compared to "other optical sensors"? Define the groups or reword for clarity

- l. 69, "continuous monitoring" is unclear given the present-day availability of automated, sub-hourly monitoring sensors. Consider using "systematic" or similar

- l. 72, what makes the site "unique"? Explain concrete reasons for uniqueness if possible, otherwise consider rephrasing

- Fig. 1, why show the glacier extent from 2016 and not the present-day? Especially since high-quality whole-glacier imagery is available (Table 1). Also, the 2022/2023 stakes installed and surveyed specifically for the present study should be clearly marked as such and distinguished from the historical archived stakes.

- l. 82, maybe it should be mentioned that the stakes were never repositioned at a fixed location, rather they were re-drilled at or close to their last location?

- l. 90, "terrestrial polar connection" yields zero results on Google Search - is the method also known by different names? Please check and possibly rephrase or provide a reference

- ll. 91-92, the two "previous studies" analyzing velocity at the site appear to be unpublished and unaccessible diploma theses. However, their findings should probably be (1) quickly presented and (2) discussed and compared with the ones of the present study.

- l. 102, "image pair velocity fields" - I think ITS_LIVE and possibly Millan rather provide annual or biennial velocity composites?

- l. 113, please provide more details about the TerraSAR-X data and methods used - at least the following: (1) acquisition dates and pairs tested, (2) software used for feature tracking, (3) tracking window sizes used, (4) any post-processing and filtering steps

- Table 1, please provide details about the "UAV" and the "Optical airborne photogrammetry" - which platform, camera, flight altitude, spatial resolution?

- l. 130, please provide more details here already about this seasonal correction - is it a single multiplication factor? How is it calculated? This information cannot be postponed to the second half of the manuscript.

- l. 150, first introduce Fig. 3 and what is displayed there, then add specification such as which stakes are included

- l. 157, the modeling of velocities should be described in detail in the Data and Methods, not just passingly in the Results. In particular, it should be explained (including formulas): (1) Where do the values for local ice thickness come from? (2) How are thickness and surface slope calculated for each stake to evolve for each year? The resulting ice velocity is a high power of both variables, thus it is highly sensitive to their precise values and variations.

- l. 163, strictly speaking, in the modeled values, we see the sensitivity of velocity to ice thickness, not directly to mass balance

- l. 165, what is "the temporally fixed choice of flow parameters"? This has not been described before, it has to be fully explained in the Methods subsection about the modeling.

- Fig. 4, this figure is very interesting. However, the "average elevation" of each stake is possibly not the most informative value here, since all elevations are clustered within less than 100 m altitude. If possible, provide also the maximum and minimum elevation of each stake (supposedly corresponding to the earliest and most recent observation dates)

- Fig. 5, to better highlight interannual changes of ice velocity, it would likely be more informative to display stake velocities not all together as absolute values, but rather as anomalies (additive or multiplicative) compared to the long-term mean at each stake

- l. 168, no data above 3000 m a.s.l. has been introduced in the text before Fig. 6 - please give a quick introduction to those stakes, are they all stakes with < 6 consecutive years of data? Are they stakes in the accumulation area?

- l. 172, these are all methodological details that belong to the Methods section, as they are necessary to understand most of the results presented so far. Also, more details are needed on the calculation of ice thickness change: was it computed only from local mass balance, fully neglecting ice advection? This would introduce a major, systematic, elevation-dependent bias in the ablation area (too fast thinning)

- Fig. 6, this interesting plot is hard to read, especially since all point measurements from all stakes are shown without distinction. At this stage, the reader does not know how many stakes are visualized on this plot, and how many stakes exist at each given point in time. Thus, it is not clear what is the time evolution here, apart from a general trend of "slowdown at all altitudes". It could make more sense to connect the points of single stakes, possibly aggregating in multi-annual intervals (even decadal aggregation) and/or excluding stakes with very few years of observation, in order to reduce the complexity of the data shown.

- Fig. 7, the thickness maps exhibit obvious major interpolation/processing artifacts, which would be strongly reflected in any calculated ice velocity. The thickness data need to be made available and/or re-examined in the light of these artifacts, if any conclusion about changes in glacier geometry is to be drawn. Also, if possible, please show simultaneous extent and thickness of the glacier rather than inconsistent dates; the 2016 extent is already in Fig. 1. Finally, the arrows indicate the flow direction, how was it determined? Please provide methodological details in the relevant section.

- l. 184, here is a description of a methodological choice and would fit very well around l. 130 if it is the same seasonality correction that is described here.

- Fig. 8, what is the standard deviation of this 30 % summer speed-up? If I understand correctly, the monthly (in summer) and annual velocities are available for each of the 8 + 4 stakes mentioned here, thus it would be quite important to show how much these stakes deviate from the 30 % estimate (and thus, how uncertain the estimate is) - especially if the correction factor is to be considered "representative and applicable" anywhere.

- l. 218, a claimed 3 cm measurement error corresponds to a good-quality dGNSS / RTK survey, whose method (instruments and protocol) should be described in the Methods section.

- l. 222, the data resolution should be introduced in the presentation of the UAV / aerial data.

- l. 229, the statistical basis of this calculation of overall uncertainty is unclear: the mean uncertainty is most definitely not the average of individual uncertainties. See e.g. [12], [13]. Another possibility would be to use leave-one-out validation of each available measurement.

- l. 233, the dataset contains maps of ice velocity and of surface slope, it should be easy to calculate proper metrics (such as a correlation coefficient) of the agreement between surface aspect and flow direction, rather than a qualitative "overall alignment of the calculated flow directions with the glacier topography".

- l. 235, there exist published methods to estimate interpolation error in areas of heterogeneous point data coverage. See for example [14].

- l. 253, is this "wide range of notable differences between the summer and winter seasons" shown anywhere? So far the Authors have presented only three spatially-aggregated, monthly data points from a single summer season (Fig. 8).

- l. 315, this conclusion can only be taken if proper state-of-the-art methods are tested and the results are shown, the well-established methods validated by several studies (including on slow glaciers) cannot be so quickly dismissed.

- l. 323, I would caution against using a map with manually tracked features as benchmark for validating other datasets, since the computed quality of such datasets would then inherit the reproducibility issues of the manually-compiled map.

- l. 329, data availability: most of the datasets mentioned and used within the study are actually missing, including the ice thickness data, any digital elevation models, and the UAV and airborne data. Only the historical stake data are provided, the other assets are rather results from the study such as the manually tracked displacements and the five-year velocity map. Unless there are legal restrictions, for both review and reproducibility purposes it is important to provide access to the actual data (not just the results) used in the study.

Bibliography

[1] https://doi.org/10.1016/j.rse.2004.11.005

[2] https://doi.org/10.5194/tc-19-219-2025

[3] https://doi.org/10.1109/TGRS.2022.3215821

[4] https://doi.org/10.1016/j.rse.2022.113038

[5] https://ethz.ch/content/dam/ethz/special-interest/baug/vaw/vaw-dam/documents/das-institut/mitteilungen/2010-2019/252.pdf

[6] https://doi.org/10.5194/gi-4-23-2015

[7] https://www.youtube.com/watch?v=PsEPqLjgj20

[8] https://doi.org/10.3189/2013AoG63A296

[9] https://doi.org/10.1016/j.rse.2013.07.043

[10] https://doi.org/10.1016/j.rse.2017.08.038

[11] https://www.persee.fr/doc/aommb_2023-1792_1893_num_1_1_856

[12] https://auckland.figshare.com/articles/conference_contribution/Computing_Uncertainty_of_Natural_Neighbour_Interpolation/9851030?file=17662652

[13] https://bgo.ogs.it/sites/default/files/pdf/bgo00423_Iurcev.pdf

[14] https://doi.org/10.1017/jog.2021.55