This study investigates glacier dynamics at VF over more than five decades (1966–2023) and provides additional insights into the spatially distributed velocity field of the recently observed period (2018–2023). Long-term in situ stake measurements provide the basis for observing historical changes in surface velocity, even well before the beginning of observations from satellite data. Furthermore, the surface mass balance (as a single value) is known for the same period of time. DEMs (digital elevation models) are available at different times. An uninterrupted timeline of the geometry can be generated using interpolation. From these DEMs, surface slope and ice thickness (based on the measured topography of 2006) can be derived on an annual basis. The methodological workflow for calculating ice thickness and surface slope is shown in Fig. . The current velocity map is generated manually after reviewing existing standard remote sensing products and techniques, which we also describe in this manuscript. A detailed overview of all used datasets (long-term and recent data) can be found in the Supplement.

A network of stake measurements has been established on VF since 1966. The primary objective of these measurements was the observation of surface ablation and thus the stakes are primarily located in the ablation area (Fig. ). Therefore, the stake positions are in general not completely representative for capturing ice surface velocities across the entire glacier. Note that the stake network was adapted over the years to capture the ablation area which changed over time due to the propagation of ablation to higher elevation in response to global warming. For this reason, the elevation range captured by the stake network varies. The locations of the stakes were measured once a year at the end of the hydrological year (end of September), together with the ablation measurements. This allows the calculation of the annual surface velocity. The measurement series are usually interrupted due to melt out of the stakes. However, the stakes are then re-drilled very close to the melt out location, so that the flow can be monitored.

Over time, different coordinate systems were used as a basis for the measurements. Meanwhile, a local system is used, which is closely aligned with the UTM (Universal Transverse Mercator), and all former measurements have been transformed into this system. Additionally, the methods for determining stake positions changed over time, starting with multiple forward intersections, continuing with terrestrial polar coordinate method, and using GNSS (global navigation satellite system) positioning for a certain period . The derived velocities have been displayed in previous studies up to the respective point in time .

In addition to the systematically monitored stake-network, 11 stakes were installed in actively moving sections of the glacier in the Taschach and Brochkogel area during the period 2022–2023. The stake positions were measured several times (4 August 2022, 10 July 2022, 21 September 2022, and 8 August 2023), allowing an analysis of seasonal variations in ice flow. VF is very likely a temperate glacier, due to its south exposure and strong melting. Therefore, the glacier is not frozen to the bed, which means that a seasonality in velocity is expected due to basal water flow. This is confirmed by water found on the glacier bed throughout a glacier core drilling on VF

A map of recent velocities on VF has been generated by compiling manual feature tracking of crevasses and stake measurements for the period 2018–2023. To achieve a standardized temporal resolution, all measured displacements were converted to a mean annual velocity in meters per year. For displacement measurements taken during the summer months (July through September) or not on equal dates in the different years (e.g. from July 2022 to September 2023) a seasonal variation needs to be considered in the conversion factor. By comparing measured summer velocities at a few locations in the Brochkogel and Taschach area with the corresponding annual average, the seasonal variation is estimated to be approximately 30 % (detailed information see Sect. ). Although the seasonal variation shows some differences between the Brochkogel and Taschach area and was only estimated for 2022/23, we consider the mean derived seasonality to be representative for the entire VF and assume its applicability to other years. Therefore, the seasonality observed in 2022/23 is applied to data covering the period from 2018 to 2023 where required. Additionally, a special set of features was created to comprise the glacier margin, where a movement of 0 myr-1 was assumed. Towards the margin, ice thickness decreases and consequently glacier dynamics slow down. Although a velocity of close to 0 myr-1 is very likely in these regions, it cannot be directly observed due to the lack of identifiable features. Therefore, specific points were defined to capture the marginal areas adequately.

In order to fill all data gaps and provide results with a uniform spatial resolution, the velocity data were interpolated onto a 50m×50m grid. Given the irregular data distribution and the presence of substantial gaps, natural neighbor interpolation was selected for its suitability in a geophysical context . Natural neighbor interpolation provides an effective balance between computational efficiency and result quality. As a localized method, it ensures that each interpolated point is only dependent on its immediate natural neighbors. A detailed explanation of the method is provided in . The technique is implemented in MATLAB using the “scatteredInterpolant” function with the “natural” interpolation method.

As previously mentioned, a detailed mass balance study indicates that the Schwarzwandzunge area behaves fundamentally different from the rest of VF . Due to this differentiation and the scarcity of current velocity data, also from the Hochvernagt area, the compiled velocity map for the period 2018–2023 of this study focuses on the Taschach and Brochkogel areas. The study region for the velocity map is shown in Fig. .

The observed datasets of surface mass balance (SMB) and in-situ surface velocities motivate an investigation of the role of glacier geometry, as variations in SMB directly affect glacier geometry, which influences surface velocities. To assess whether the observed velocity trends can be reproduced by geometric changes, we apply a simplified shallow-ice approximation (SIA) method. This approach requires only the geometry, consisting of the surface slope (α) and the ice thickness (H), to estimate the velocity e.g.. The SIA approach is also referred to as a “zero-order” model, where only vertical shear stress gradients are taken into account. We therefore do not regard this as ice dynamic modeling in the strict sense, but rather a first approximation. The selected physical flow parameters are summarized in Table . The total surface velocity (utotal) consists of the deformation velocity (udeform) and the basal sliding velocity (ubasal): 1utotal=udeform+ubasal

For calculating the deformation velocity soft and temperate ice was assumed by choosing a rheology parameter A corresponding to a temperature of 0 °C . The deformation velocity is given by: 2udeform=2An+2τnH, where the shear stress τ is described as: 3τ=-ρgHsin⁡(α)

The basal sliding velocity estimation requires a basal sliding coefficient C. As a priori determination is challenging, we tuned the coefficient so that the modeled velocity fall within the range of the measured velocities. The basal sliding velocity can be described as: 4ubasal=C⋅τn

In order to apply the SIA approach, annual ice thickness and surface slope are required. For this purpose, DEMs were generated by interpolation between available DEMs at known times (; and more recent maps published by the group at BAdW) for workflow, see Fig. . All available datasets are listed in the Supplement.

Instead of assuming linear interpolation over time, the interpolation between two consecutive DEMs was scaled according to the cumulative SMB. The annual mass balance is determined through the glaciological method and is described by a single glacier-wide value per year, as area distributed annual mass balances from the glaciological method are not available for all periods . For example, the DEM from 1969 and 1982 is available, the years in-between are interpolated. Therefore, we started to sum the SMB during this period of time and used as relative measure of surface elevation change. This means, that the elevation change between the DEMs from 1969–1982 corresponds to 100 % of the cumulative SMB from 1969–1982 (in this example -596 mm w.e.). The yearly change in DEMs were generated according to the cumulative SMB reaching the respective year, for example cumulative SMB 1969–1976 with -520 mm w.e. corresponds to 87 %, meaning that 87 % of the elevation change from 1969–1982 would be applied to the DEM in 1976. In this way, elevation changes are distributed non-linearly in time and reflects periods of higher or lower loss/gain. The error of non-linear ice transport over time is neglected, as the velocities are relatively low. The respective ice thickness is derived by subtracting the bed topography of 2006 from the corresponding interpolated DEM. Surface slope was calculated from the annual DEM using five grid cells (k=5) in each horizontal direction, with a grid size (dx) of 20 m, the baseline length (L) is 200 m. The resulting ice thickness and surface slope were restricted to the last known glacier extent. To compare the calculated velocity to in-situ stake observations, we extract the calculated velocity from the 20m×20m grids using a nearest-neighbor approach.

The examination of historical velocity data is essential to assess long-term trends and investigate the development of glacier dynamics. For this purpose, the stake network data, which has been measured annually at VF since 1966, is analyzed. Figure  displays the stake positions during the years. Since the focus lies on evaluating the temporal evolution of the ice movement, only time series of at least six consecutive years are considered. The data clearly demonstrate that stake displacements follow the flow structures of the different glacier tributaries. In addition, the generally decreasing distance between the annual stake positions of each data series indicates that stake velocities decrease over time. Figure  presents the annual velocity of six particularly long time series, highlighting a brief acceleration phase around 1980, followed by a pronounced and sustained decline. This velocity peak concurs with a period of positive mass balance, while the subsequent decline coincides with a sustained negative mass balance since 1980. This clearly illustrates the sensitivity of the ice movement to changes in surface mass balance.

To evaluate a possible reproduction of the velocity trends using only geometry, we compared the measured velocities with modeled velocities (from the SIA approximation) in Fig. . While the modeled values provide the same dynamic trends as the measured, offsets occur, which can be attributed to uncertainties in the input data (especially the ice thickness) and the temporally fixed choice of flow parameters. However, it is remarkable that even the small effect of slightly positive mass balance years on the glacier geometry, results in very pronounced changes of ice flow.

Over time, surface velocities at all altitudes decrease significantly (see Fig. ). However, for altitudes exceeding 3000 ma.s.l. in the previous years, this trend cannot be conclusively established due to insufficient data availability. In the range of 2900–3000 ma.s.l. the reduction is most evident. In addition, Fig.  shows the regional distribution of surface velocities in connection with the respective ice thicknesses. Overall, a decrease in velocities can be observed at all altitudes, although their magnitude varies. The decrease in surface velocities is particularly pronounced at the Schwarzwand tongue between the periods a (1970–1980) and b (1990–2000). In contrast, the decline is less significant in the area of the Taschach tongue. These differences can be attributed to the interaction between ice dynamics and ice thickness. Between the two periods, the ice thickness reduction at Schwarzwand tongue is substantially greater, being approximately 35 %, whereas at Taschach tongue, the reduction was only about 10 %.

To quantify the ice dynamic process, we performed an analysis based on the observed geometry change, using surface elevation change derived from known DEMs (hereafter referred to as surface elevation change). Based on the ice thickness equation , the dynamic contribution to ice thickness change can be computed by subtracting the local surface mass balance (SMB) from local changes in ice surface elevation. The residual component represent the local dynamic contribution to the vertical surface adjustment (assuming a stationary bedrock). The spatially distributed SMB for VF are available only from 1996 ongoing (earlier information as glacier-wide values). With a spatially distributed SMB and a surface elevation change map, the difference can be calculated as local dynamic contribution to elevation change.

We apply this analysis to the period 2016–2018, for which data availability is particularly good. Although this is a very short period with regards to glacier response times, it is assumed to be representative of the current ice dynamics, as the glacier already has very low velocities during this period, similar to those observed in 2018–2023. The cumulative SMB for the years 2017 and 2018, is shown in Fig. a, while the surface elevation change between the years 2018 and 2016 can be seen in Fig. b. The difference between the two maps provide the local dynamic contribution to elevation change (Fig. c).

In the accumulation areas (or former accumulation areas), negative values (red areas) indicate that the ice transport has a vertically downward component. In the ablation areas, on the other hand, positive values (green areas) indicate vertical mass compensation through ice transport. Figure c shows that the Taschach and Brochkogel areas exhibit hardly any vertical mass compensation (values very close to 0). In contrast, a local dynamic contribution to elevation change is still present in the former accumulation areas of these two sub-areas. The two green areas in the Taschach area provide good control. These are rock islands that were not excluded from the SMB map. However, the rock islands have remained stable during the period and have not moved, which is why they appear incorrectly as dynamic contribution to elevation change.

A closer look at the Schwarzwand area in Fig. c reveals a stronger dynamic contribution to elevation change. In the northern accumulation area of the Schwarzwand area (as well as on the Hochvernagt plateau), positive values are observed that cannot be plausibly attributed to a dynamic contribution to elevation change. It seems that a substantial accumulation area is missing. However, in the orthoimages, these areas are partially covered by snow and have hardly any features. It can therefore be assumed that the surface models in this area are uncertain. In addition, the SMB map (Fig. a) is largely interpolated (in particular in the northern part of Schwarzwand area), further increasing the uncertainty of the estimated dynamic contribution to elevation change in this region. It is also possible that these are formerly significant accumulation areas that are no longer very pronounced. Pronounced negative error values are present in the northern part of the Hochvernagt plateau (Fig. c). These originate from the surface elevation models, where an error has likely occurred.

In order to enable a numerical estimation of the processes, which excludes outliers as far as possible, the median of the absolute mass balance and the absolute dynamical contribution were calculated separately for the three glacier parts. The median of the absolute mass balance amounts to 1.48, 1.32, and 1.06 myr-1 for the Taschach, Schwarzwand and Brochkogel area, respectively. The corresponding median of the absolute dynamical contribution to elevation change are 0.54, 0.43, and 0.74 myr-1. For the mass balance determined by the glaciological method, a total uncertainty of less than 0.40 myr-1 can be assumed . For the period 2016–2018, an uncertainty of approximately 0.28 myr-1 can therefore be calculated, taking into account the law of error propagation. The mean errors of the DEMs can be estimated at approximately 20 cm for 2016 and 10 cm for 2018 . Taking into account the law of error propagation, this results in an uncertainty of 0.11 myr-1 for the surface difference from 2018 to 2016. For the dynamic contribution to elevation change, calculated as the difference between SMB and surface difference, a total uncertainty of approximately 0.30 myr-1 can be assumed. Overall, further uncertainties must be taken into account, such as the compression and thus the change in altitude of possible firn areas due to a change in density, especially in the accumulation areas.

The measured values at different times (Fig.  and Table ) clearly indicate that ice surface velocities are subject to seasonal variation. This is based on stake measurements in active moving terrain in the period 2022–2023, located in the Brockogel (4 stakes) and Taschach area (8 stakes). In the Brochkogel area, summer velocities are around 20 % (July–August) and 49 % (August–September) higher than the respective annual values. In the Taschach area, summer velocities exceed the corresponding annual mean by approximately 30 % (July–August) and 31 % (August–September). Thus, the seasonal variation averages 30 %. However, we consider the derived seasonality to be representative for the entire VF for July–September and assume its applicability to other years, although 2021/22 was a year with an extremely negative mass balance.

This chapter describes the results of the uncertainty analysis, details of the error calculation can be found in in Appendix . The accuracy of all stake positions depends on the respective measurement method (multiple forward intersections, terrestrial polar coordinate method, and GNSS (Global Navigation Satellite System) positioning). In all cases, the placement of a prism or Antenna, depending on the method, requires a small horizontal offset due to the physical presence of the wooden stakes, which prevents exact alignment with the center of the borehole. Moreover, if the boreholes are not drilled vertically in the ice, melting can also cause further displacement biases. In combination with the measurement error, which is in the order of 3 cm, these factors result in an estimated general uncertainty of ±10 cm. Therefore, all positions in the stake data set, as well as in the figures of the multi-year stake displacements, are affected by this uncertainty. On average, seasonal variation is 1.30 (130 %), with a standard deviation across all observations of 0.37 (37 %). In addition to stake measurements, manually tracked surface features (e.g. crevasse intersections) were used to derive independent glacier surface velocities. The uncertainty of localizing these features depends on the image resolution (≤20 cm pixel size), the feature size (≥40 cm), possible sideways oblique views of the features (depending on the respective angle, in the order of around 5 cm) and the co-registration accuracy of the images. Considering these parameters, the feature position uncertainty can be estimated at ±35 cm. Manual case-by-case verification and explicit searching for crevasse intersection points ensure that only features with minimal change due to ablation are selected, thus guaranteeing this very high quality.

A covariance propagation law can be used to estimate the uncertainty of the calculated velocities, depending on the temporal resolution and the data detection method. The resulting velocity uncertainties represent one standard deviation (1σ) and are listed in Table , a detailed description to the calculation is in Appendix . It is also demonstrated that an approximate total measurement uncertainty of 0.31 myr-1 can be estimated. For instance, feature tracking with a 2 year interval results in a velocity uncertainty of ±0.25 myr-1, whereas a 1 year interval yields ±0.49 myr-1.

The natural neighbor interpolation of all observations listed in Table  yields a correlation coefficient (R) of 0.94 for eastward velocity and 0.95 for northward velocity. The residual variance is 0.15 m, respectively, which indicates a robust and consistent representation of glacier flow. To examine the uncertainty of the interpolation more closely, a leave-one-out method was performed (e.g. . This shows that there is no significant dependence of the misfit (between interpolated velocity and measured velocity) and the distance to the next observation. However, a relative interpolation uncertainty of 57 % and an absolute interpolation uncertainty of 0.8 myr-1 can be estimated from the standard deviation. Detailed calculations and explanations can be found in Appendix . The reliability of the interpolated map is further supported by the overall alignment of the calculated flow directions with the glacier topography (slope direction), with no significant outliers observed. The median angle deviation is 43°, but larger deviations may occur at very low magnitudes. Despite the estimated interpolation error, the velocity map becomes less reliable in regions with limited data availability. In zones where observational data is lacking, there is insufficient information, making the reliability of the velocity map for understanding ice flow dynamics in these areas questionable. For the period 2018–2023, annual surface velocities are assumed to be constant. This assumption is supported by stake measurements, which show a maximum year to year variation (decrease) of 10 % within measurement uncertainty. We therefore assume that there is no significant change during this period.

This subsection discusses the observed changes in glacier flow. Long-term and seasonal variations are interpreted in relation to changes in ice thickness and mass balance. When examining historical data, a general decrease in glacier velocity over time is prominent. This trend can be attributed to two main factors. First, the stakes slow down as they approach the glacier tongue, in line with the general velocity decrease towards the tongue. Second, the reduction in ice thickness due to the continuous negative glacier mass balances influences the velocities, as deformation rates are directly dependent on ice thickness. Driving stress is directly related to ice thickness and thus a reduction in ice thickness leads to lower driving stress and therefore to a reduction in deformation rate. On VF the magnitude of velocity change varies depending on the region. This can be seen particularly at Schwarzwand tongue, where both the ice thickness and the velocities decrease significantly, even more than on the rest of VF. As ice thickness decreases, deformation rates and velocities decline, a well-documented pattern in glaciological studies. One exception to the general slowdown is evident, as marked by three black circles in Fig. . These cases show a temporary increase in velocity during the period 1975–1985. The accelerated ice flow during this period likely originated from the temporarily positive mass balances and an accompanying small glacier advance around 1980 . Neighboring glaciers, such as Hintereisferner show a similar velocity peak around 1980, followed by a decrease . The peak aligns with a wider regional trend, as a significant proportion of glaciers in the European Alps experienced advances during this temporarily relatively cool phase . We have found evidence that the displacement of ablation stakes at VF react quickly to a change in the mass balance. An ongoing period of predominantly negative mass balance starting in the early 1980s marks the onset of a decrease in surface velocity in the ablation area. Ice thicknesses are rather small in the tongue areas of VF, where the stake measurements are carried out. Ice velocity is very sensitive to changes in ice thickness and the relative change of ice thickness due to strong ablation. With already reduced dynamic compensation leads to a fast reduction in ice velocity. In a year with a strongly negative mass balance, the strong melt in the ablation area directly decreases ice thickness, which represents the current status of the glacier. have also found evidence that ablation stakes can be well suited to reflecting the current status of a glacier. They show that the Kesselwandferner (also in the Ötztal valley) shows relatively direct response of ablation area ice dynamics to changes in mass balance, regardless of the geometry of the glacier. On VF, there is insufficient data to substantiate this adequately. Further studies (including in other areas) are necessary to investigate this effect in more detail. Comparable behavior has been reported from long-term observations of the d'Argentière glacier, Mont Blanc area. show an ongoing negative net mass balance since 1982, with a direct reflection (delayed by a maximum of 3 years) in the velocity. They describe that the change in velocities in the upper part of the glacier is smaller than in the lower part (ablation area). The velocity trend thus fits the VF timeline quite well.

A quantitative analysis of recent geometry change (period 2016–2018) at VF indicates that SMB is the dominant control, while local dynamic contribution to elevation change remains secondary. When interpreting the contributions of SMB and local dynamic contribution to elevation change, various uncertainties must be taken into account. For the investigated period (2016–2018), measurement uncertainties arise from SMB estimation (0.28 myr-1), DEM differencing (0.11 myr-1) and the derived local dynamic contribution to elevation change (0.30 myr-1). In addition to these measurement uncertainties, various physical processes may influence the observed elevation changes. For example, snow cover in the accumulation area can affect the DEM accuracy, while small local wind-driven snow transport events can modify the accumulation pattern and thus SMB. Additionally, avalanches and debris covered parts of the glacier may contribute to elevation change or altered ablation, although their overall impact at VF is assumed to be limited. These processes are not considered in the applied approach and may therefore lead to additional variations in the estimated local dynamic contribution to elevation change, particularly on small spatial scales. Despite these uncertainties, the significantly higher magnitude of SMB remains the dominant control on recent elevation change. The median absolute values for SMB (Taschach: 1.48 myr-1, Schwarzwand: 1.32 myr-1 and Brochkogel area 1.06 myr-1) and the local dynamic contribution to elevation change (Taschach: 0.54 myr-1, Schwarzwand: 0.43 myr-1 and Brochkogel area: 0.74 myr-1) suggest that the ice dynamic is not able to fully compensate the losses due to melt, especially in the ablation area. There are indications that there may be slightly more compensation of elevation change through ice dynamics in the Schwarzwand area and that this area behaves fundamentally different from the rest of VF. This was already suggested by and can be confirmed here. Figure  also suggests a higher dynamic on the Schwarzwand tongue compared to the other tongues. Due to a lack of data (no spatially distributed SMB before 1996), it is not possible to quantitatively assess these processes for the early years of glacier observation. However, it can be assumed that the local ice dynamic contribution to elevation change was more pronounced during this time, as the measured velocities were significantly higher, particularly around 1980, with an absolute SMB being significantly smaller than it is today.

Short-term seasonal variations play a role in the general glacier flow behavior. A significant temporal variability is observed, with a notable difference between the summer and winter seasons. There is general agreement that this variability is influenced by the glacier size, as well as internal and subglacial hydrology, although this has not been clarified conclusively and is still a subject of current research . In general, the seasonal variability can be rather large, with summer velocities either comparable or even up to 300 % higher than winter values . At VF, summer velocities are approximately 30 % higher than the annual mean, with a standard deviation across all observations of 37 %. The relatively high uncertainty results from the fact that the absolute measured values over a month (July–August or August–September) have maximum values of about 50 cm. However, these values have a relatively high measurement uncertainty of ±14 cm. Our measurements were conducted in summer 2022 in an area of higher ice flow, which is relatively high-elevated, but it is still not part of the accumulation zone and is strongly affected by melting. Meltwater input and the related increase in subglacial water pressure can reinforce the basal sliding of the glacier during the summer until an efficient drainage system is developed . Crevasses and moulins can encourage the meltwater input to the glacier bed, increasing the basal sliding . At our measurement area, such an increase in meltwater input can occur by existing crevasses, potentially also from upstream regions. As a result, seasonal variation may be stronger in our measurement area, particularly as efficient drainage systems are probably only well developed further down on the glacier tongue. Whether this summer velocity increase is representative for larger areas of VF and for other years has yet to be demonstrated. No assessment can be made regarding the time of highest seasonality because of the lack of measurement data, other studies refer to spring , due to high melting rates combined with a not yet developed drainage system.

This part outlines the current challenges posed by very low ice velocities in combination with pronounced surface ablation. The annual surface velocities at VF are on average around 1 myr-1 while maximum velocities reach around 4 myr-1. These low velocities, in combination with strong surface ablation, pose a challenge for measuring the glacier flow using remote-sensing techniques like feature tracking. A sufficient displacement of the features between the consecutive acquisitions is needed. The retreat dataset is based on consecutive Sentinel-1 SAR acquisitions with minimum temporal baselines ranging between 6–48 d, depending on the region, and a maximum of 96 d . For the Alps, a minimum temporal baseline of 48 d was employed. However, considering a spatial resolution of Sentinel-1 IW data of 10 m, the displacement rate at VF would correspond to 0.025 pixels for 96 d between the data takes, which is too small to be accurately captured by feature tracking. Typically, upsampling rates of 2–8 are employed to obtain sub-pixel displacement rates. However, the displacement rates at VF are too low, even for high-resolution TerraSAR-X stripmap data (2 m resolution). Moreover, the accumulation area has decreased significantly in recent decades, leading to widespread surface melting across large parts of the glacier . The magnitude of point mass balance due to melting exceeds horizontal velocities in many areas on VF, meaning that observed feature velocity is heavily overlaid by surface changes caused by melting in summers, further limiting the unambiguous identification of features between the consecutive acquisitions. Interferometric techniques, such as SAR coherence tracking or InSAR displacement measurements, are hampered as well by the strong ablation rates during summers and snowfall throughout winters. These climatic factors also explain the observed limitations in the tacking results when using long temporal baselines, as tested for the TerraSAR-X data and selected ITS_LIVE image pair velocity fields. Moreover, the limitations in the individual image pair results can not be compensated by combining multiple results in monthly or yearly mosaics, since all output is affected by the various issues (see Figs.  and and Sect. ). Overall, the limited displacement rates, high surface melt during summers plus high accumulation during winter as well as the irregular distribution of surface features impede the capturing of the ice flow at VF by standard remote-sensing techniques. For other relatively slow-flowing glaciers, such as the Griesglacier, surface velocities could be derived over short time periods using UAV-based imagery and software such as IDMatch, a tool for automated velocity derivation . Compared to the VF, the average velocities on Griesglacier with more than 1 m in just 1 month are significantly higher than the average velocities at the VF of about 1 myr-1. Furthermore, a short test of the IDMatch software using the UAV VF datasets from July 2022 and September 2022 results in no correlation between the images. Regardless of the software and the potential detection of identical features, surface changes caused by strong ablation at VF remain a dominant source of uncertainty, as illustrated in Appendix . We do not want to rule out the possibility of automated derivation, but it would be essential to additionally correct for the false values caused by the ablation-induced change in features (a change not caused by ice flow). For this reason, we have chosen manual feature tracking.

Our work reveals that standard remote-sensing velocity products, including those evaluated in this study, have limitations when applied to a slow-flowing glacier with pronounced surface ablation. Due to these limitations, ground-truth measurements and/or manual feature tracking can be the method of choice, although the methods are very demanding compared to standard remote sensing techniques. The manual feature tracking enables a focused segmentation of high-quality points, a task that remains challenging even to the human eye due to the significant surface changes over time (see Appendix ). It allows simultaneous plausibility checks of individual points, ensuring greater reliability.

Although we were not able to generate a surface velocity map for the entire VF, we successfully constructed a valid map of current velocity of the Taschach and Brochkogel area, demonstrating key dynamical mechanisms operating on the glacier. The map integrates multiple datasets for the period 2018–2023, primarily stake measurements and manually tracked crevasse features. Consequently, areas with many crevasses are represented by numerous data points, whereas regions lacking distinct surface features are sparsely covered, where stake measurements remain the only basis for the interpolated velocity map. A clear correlation is evident between the appearance of crevasses, indicated by areas with a high density of data points, and high velocities, up to 4 myr-1. A connection of crevasses and velocities is well-documented in the literature, including detailed studies such as . The assumption of zero velocity at the glacier margin when generating the velocity map neglects possible transverse stress couplings. Nevertheless, the dynamics at the margin of the glacier are significantly slower than in the center. Since the VF itself flows relatively slowly even in the center, the velocities at the margin will be close to 0 myr-1.