the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 06 Aug 2026
High-resolution spatio-temporal variability of surface albedo on the glaciers of Hurd Peninsula, Livingston Island (2018–2025): controls by snow metamorphism, surface impurities and terrain roughness
Abstract. Surface albedo is a primary control on the energy balance of maritime Antarctic glaciers, yet its fine-scale spatial and temporal variability remains poorly constrained. Here we present a multi-campaign broadband albedo dataset acquired on Hurd Peninsula (Livingston Island, South Shetland Islands) during four austral summer field seasons (2018, 2019, 2024, 2025), using a portable albedometer mounted on a snowmobile traverse at 5 s sampling resolution. Under clear-sky conditions, surface albedo undergoes exponential decay described by α(t) = α0e−kt with k = 2.24 × 10−5 5 s−1 (R2 = 0.98), driven by wet-snow metamorphism and confirmed independently by time-lapse microscopy and field spectroradiometry. A statistically significant positive interannual trend in surface albedo (+0.016 yr−1, R2 = 0.57) is documented over the study period, most pronounced at elevations below 200 m above sea level (a.s.l.), and attributed to increased summer snowfall frequency associated with the regional intensification of precipitation. Residuals between observed and modelled albedo reveal two distinct classes of spatial forcing: (i) biological and mineral impurities — Chlamydomonas nivalis blooms recurrently concentrated below 100 m a.s.l. and cryoconite deposits near 250 m a.s.l. — and (ii) terrain roughness and slope, whose correlation with albedo residuals reverses sign between years of contrasting snow cover. An integrated albedo–altitude profile (n = 4,219) identifies the 200–260 m a.s.l. band as the locus of maximum variability, coinciding closely with the mean equilibrium-line altitude (ELA) of Hurd Glacier (∼203 m a.s.l.). These results demonstrate that broadband albedo on maritime Antarctic glaciers cannot be adequately characterized by temporal decay models alone, and that impurity distribution, terrain geometry, and proximity to the ELA must be explicitly accounted for in energy balance and remote sensing applications.
- Preprint
(11828 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-4094', Anonymous Referee #1, 08 Sep 2026
-
CC1: 'Reply on RC1', Susana del Carmen Fernández Menéndez, 27 Sep 2026
Response to Reviewer 1
We thank the reviewer for the careful and constructive assessment. We agree that several statements require clarification or moderation, particularly concerning temporal inference, the transferability of the stationary decay experiment, and the interpretation of the reference-curve deviations. Our responses and proposed revisions are summarized below.
- Temporal variability
Reviewer concern: Four campaigns are insufficient to establish a long-term albedo trend, particularly given the influence of summer snowfall.
Author response: We agree that four field campaigns cannot, by themselves, support inference of a long-term temporal trend. Long-term and seasonal albedo variability over Hurd Peninsula has already been investigated by the authors using MODIS time series and ground observations (Calleja et al., 2021; Calleja et al., 2024). The objective of the present study is different and complementary: to resolve the metre-scale spatial variability and sub-hourly evolution of surface albedo that cannot be captured by satellite products.
The limited extent and coastal configuration of the Hurd Peninsula glacier system restrict MODIS coverage to only nine pixels, approximately half of which are mixed pixels containing variable contributions from glacier ice, exposed land and ocean. MODIS is therefore suitable for characterizing seasonal and multiannual albedo evolution, but it cannot resolve the effects of local surface impurities, terrain roughness, elevation gradients or rapid snow metamorphism addressed by the present field measurements.
We will therefore remove the interpretation of the four campaigns as a long-term trend and describe the differences as campaign-to-campaign variability. The previously published MODIS analyses will provide the long-term temporal context, whereas the present study will focus on high-resolution, spatially distributed and sub-hourly in situ variability. We will clarify this distinction in the Introduction, Discussion and Conclusions.
- Stationary decay experiment
2.1 Experimental conditions
Reviewer concern: The date, time, illumination conditions and surface-energy-balance context of the stationary experiment are insufficiently described.
Author response: We agree that further information is required. The experiment was conducted on 14 January 2024, at approximately 200 m a.s.l., under clear-sky conditions close to local solar noon. We will provide the exact acquisition interval and the available meteorological and incoming-radiation information.
We will also clarify that not all components of the surface-energy balance were measured. Consequently, the experiment cannot isolate the contributions of radiative, turbulent and conductive fluxes, or exclude other surface processes. The suggested references concerning weathering-crust development will be considered in this context.
2.2 Transferability
Reviewer concern: A function derived from one 40 min experiment may not be applicable to all mobile transects.
Author response: We agree that one experiment cannot define a universally applicable decay coefficient for Livingston Island or maritime Antarctic glaciers. We will therefore describe the function as a site- and condition-specific empirical clear-sky reference curve, rather than as a universal physical parameterization.
The accompanying microscopy and spectroradiometry document simultaneous microstructural and optical evolution during the experiment, but they do not demonstrate that the same coefficient applies under all meteorological and surface conditions. This limitation will be stated explicitly.
The four transects included in the study were selected from a larger field dataset because they were acquired under clear and comparatively stable illumination. A complete traverse requires approximately two to three hours, and such stable conditions are uncommon on Livingston Island. Measurements acquired under overcast or rapidly variable cloud conditions require a different treatment because changes in incoming irradiance and in the direct-to-diffuse radiation ratio may dominate the temporal signal.
- P95 normalization
Reviewer concern: P95 is calculated from the complete traverse and cannot be interpreted as an albedo value physically measured at .
Author response: During method development, we evaluated the maximum, arithmetic mean, median and P95 as alternative definitions of the campaign-specific reference amplitude. The exponential formulation was retained under all four alternatives. P95 was selected for the principal analysis because it provides a robust estimate of the upper envelope of each traverse without relying on a single extreme observation, as occurs with the maximum, and without being displaced towards the central tendency of a spatially heterogeneous series, as occurs with the mean and median. In this sense, P95 was used as a mathematical normalization device rather than as a surface-albedo value physically observed at .
We nevertheless agree that the consequences of this methodological choice should be documented explicitly. We are therefore willing to include the results obtained using the maximum, arithmetic mean, median and P95 as a sensitivity analysis in the revised manuscript. This comparison will report the corresponding changes in bias, MAE and RMSE and, where relevant, in the terrain relationships and spatial distribution of the principal negative-albedo anomalies. Because is held fixed, preservation of the exponential form is mathematically imposed and will not itself be presented as evidence of robustness; instead, robustness will be assessed from the stability of the downstream results under the alternative normalizations.
- Reference-curve deviations
Reviewer concern: The interpretation of the residuals and their use in the terrain and impurity analyses are unclear.
Author response: We agree that model residuals implies more physical certainty than is justified. We will refer to these quantities as signed deviations from the empirical reference curve.
These deviations do not uniquely separate temporal and spatial forcing. They may reflect surface state, illumination geometry, terrain, impurities, meteorological variability and measurement uncertainty. Relationships with slope and TRI will therefore be presented as conditional spatial associations rather than as evidence of a unique causal mechanism.
Likewise, a negative deviation alone will not be used to identify impurities. Potential impurity-related anomalies will only be discussed where the spatial signal coincides with independent field observations, microscopy and/or a distinctive spectral response.
- Representativeness
Reviewer concern: The representativeness of the stationary experiment and the regional applicability of the results require further consideration.
Author response: We agree. The experiment represents one specific clear-sky, wet-snow situation and cannot be considered climatologically representative of Livingston Island. Numerical coefficients, elevation thresholds and effect magnitudes will therefore be interpreted as specific to the site and measurement conditions.
The objective of the study is not to establish a regional parameterization, but to provide a high-resolution characterization of albedo variability within the Hurd Peninsula glacier system. The transferable contribution is the field protocol and analytical framework, not necessarily the numerical values obtained at Hurd Peninsula.
- Surface impurities
Reviewer concern: The sampling information is insufficient, snow and ice surfaces are not consistently distinguished, and species-level identification is not fully supported.
Author response: We agree that species-level identification is not supported by our observations. Taxonomic characterization was not an objective of this albedo study; the biological material was included only as a potential light-absorbing surface impurity. Recent molecular work from Charrúa Glacier, Livingston Island, identified Sanguina nivaloides as the dominant taxon in red-snow samples, supporting the reviewer’s concern regarding our use of Chlamydomonas nivalis. We will therefore replace Chlamydomonas nivalis with the non-taxonomic term red snow algal assemblages, add the relevant regional references, and distinguish explicitly between observations on snow and on bare ice. No species-level identification will be claimed.
- Figures and methods
Reviewer concern: Several methodological and graphical details require clarification.
Author response: We will:
- Separate the stationary temporal experiment from the mobile along-track measurements in the graphical abstract.
- Clarify that elapsed time and along-track distance are not interchangeable variables.
- Describe the albedometer mounting and levelling procedure and report the associated uncertainty.
- Identify the source and characteristics of the background image in Fig. 1.
- Add travel-direction arrows and start/end points to the traverse maps.
- Remove the repeated GNSS description.
- Move interpretive statements from figure captions to the Results.
- Reduce repetition between the text and Table 5.
- Correct the Fig. 8 caption.
- Clarify the distinction between the predominantly ablation-zone sector below 200 m a.s.l. and the higher-elevation sector, while acknowledging temporal variability in the ELA.
- Moderate the interpretation of the reduced albedo observed above approximately 270 m a.s.l.
Remote sensing will be presented as complementary to the field measurements. Satellite products provide regional and seasonal coverage, but they cannot resolve the same metre-scale relationships among albedo, topography and localized surface conditions.
- Supplementary material
Reviewer concern: Figure A1 and Table A1 could not be located from the main text.
Author response: Figure A1 and Table A1 are included in the supplementary material, but we agree that they were not adequately cross-referenced. We will add explicit citations at the relevant locations in the manuscript and verify that the supplementary files are correctly linked and publicly accessible.
- Data and code
Reviewer concern: Only the 2025 dataset appeared to be accessible, and the GitHub link did not lead to a public repository.
Author response: The ordered georeferenced tracks are deposited in the Spanish National Polar Data Centre (CNDP). Their temporal sequence defines the direction of travel and corresponds to the elapsed-time axis used in Fig. 5. However, the reviewer’s difficulty accessing the files indicates that the links or catalogue structure were not sufficiently clear.
We will verify the public accessibility of the datasets for all four campaigns and provide direct links in the Data Availability statement. If access through the CNDP interface remains problematic, we will also mirror the ordered track files in a dedicated public GitHub repository. The CNDP will remain the primary long-term archive, while GitHub will provide a direct alternative access route.
We will also replace the general GitHub profile link with a direct link to a dedicated public repository containing the analysis scripts and the necessary processing information.
Closing statement
We thank the reviewer for the constructive comments. In the revised manuscript, we will clarify the scope and limitations of the temporal analysis, present the exponential function as a condition-specific empirical reference, and document the sensitivity of the results to the alternative normalization choices. These revisions will strengthen the interpretation without altering the main contribution of the study: a high-resolution, spatially distributed in situ characterization of surface-albedo variability across the Hurd Peninsula glacier system. We welcome further discussion during the open-discussion stage
References
F. Calleja et al., "Snow Albedo Seasonal Decay and Its Relation With Shortwave Radiation, Surface Temperature and Topography Over an Antarctic Ice Cap," in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 14, pp. 2162-2172, 2021, doi: 10.1109/JSTARS.2021.3051731. keywords: {Snow;Temperature measurement;Surface topography;MODIS;Ice;Antarctica;Temperature sensors;Antarctica;digital terrain model;remote sensing;shortwave radiation;snow albedo;snow metamorphism;temperatura)
Calleja, J.F.; Muñiz, R.; Otero, J.; Navarro, F.; Corbea-Pérez, A.; Reijmer, C.; de Pablo, M.Á.; Fernández, S. Spatiotemporal Evolution of the Land Cover over Deception Island, Antarctica, Its Driving Mechanisms, and Its Impact on the Shortwave Albedo. Remote Sens. 2024, 16, 915. https://doi.org/10.3390/rs16050915
Citation: https://doi.org/10.5194/egusphere-2026-4094-CC1
-
CC1: 'Reply on RC1', Susana del Carmen Fernández Menéndez, 27 Sep 2026
-
RC2: 'Comment on egusphere-2026-4094', Anonymous Referee #2, 28 Sep 2026
Review of the manuscript entitled “High-resolution spatio-temporal variability of surface albedo on the glaciers of Hurd Peninsula, Livingston Island (2018–2025): controls by snow metamorphism, surface impurities and terrain roughness”
The introduction is comprehensive and clear, but it creates some confusion on two levels. On the one hand, it provides a general overview of studies on albedo and its temporal evolution in glacial contexts; on the other hand, it attempts to describe specifically what is happening on the Antarctic Peninsula and at the sites involved in this study. These two levels are not always clearly separated, and I would try to clarify this.
The description of the field measurement protocol should be expanded. It is currently not clear whether albedo measurements were acquired continuously while the operator was moving along the glacier transects, or whether measurements were taken at stationary points at 5 s intervals. This distinction is important, as measurements collected while walking may be affected by changes in sensor orientation, height above the surface, operator position and body shadow, as well as by synchronization between irradiance and GPS measurements. Please better describe how the instrument was carried, whether the operator stopped for each measurement or measurements were acquired while moving, how sensor orientation and height were maintained, and how GPS position was associated with each irradiance measurement.
I have concerns about the temporal normalisation applied to the mobile traverse data. The exponential decay coefficient is derived from a single 40-minute stationary experiment performed at one location, on one day in January 2024, under clear-sky conditions, but the same coefficient is subsequently applied to four different field campaigns (2018, 2019, 2024 and 2025). It is not clear to me that such a decay rate can be assumed to be invariant across years, locations and snow conditions, since snow metamorphism depends on the initial snow state and on meteorological and surface-energy conditions.
I am also not fully convinced by the definition of the campaign specific α0 as the 95th percentile of each traverse. The P95 may identify some of the highest-albedo surfaces encountered during a traverse, but why should this value represent the albedo of the entire traverse at t = 0? In a spatially heterogeneous glacier surface, high values may simply reflect locally cleaner or fresher snow, different grain properties, impurities or terrain effects, and the observation corresponding to P95 may have been acquired well after the beginning of the traverse.
More generally, the stationary experiment simultaneously captures changes in snow properties and changes in illumination geometry, if I understood the setup correctly. Although spectral measurements provide evidence that snow metamorphism occurred, it is not obvious that the entire observed broadband decay can be uniquely attributed to metamorphism. What about the effect of changing solar geometry? This is important because the residual analysis assumes that, once the exponential temporal component has been removed, the remaining variability is spatial. Any mismatch in the temporal model would therefore propagate directly into the residuals. I suggest that the authors better justify this assumption, explore the sensitivity of their results to the choice of k and α0, and discuss the limitations of applying a decay function derived from a single short experiment to the complete multi-year dataset.
Other concerns relate to the spatial scale represented by the DEM-derived variables, particularly TRI. REMA has a spatial resolution of 8 m, while several surface features mentioned as examples of “irregular microtopography”, such as crevasses and supraglacial drainage features, may have dimensions of only a few metres and therefore be unresolved or strongly smoothed in the DEM. Moreover, because TRI is calculated using the central cell and its eight neighbouring cells, it represents topographic variability over a scale of several tens of metres rather than microtopography at the scale of individual albedo measurements. The scale mismatch between near-point radiometric observations and terrain variables seems important. I suggest clarifying the spatial scale actually represented by REMA-derived slope and TRI and discussing this limitation, particularly because weak correlations with TRI may partly result from the inability of the DEM to resolve the relevant surface roughness.
One of my major concerns is the interpretation of the interannual analysis as evidence for a temporal trend. Although the study period spans 2018-2025, the dataset includes only four independent field seasons (2018, 2019, 2024 and 2025), so the effective temporal sample size for assessing an interannual trend is n = 4. If the reported linear trend (+0.016 yr-1, R2 = 0.57) is based on the four annual means, I do not see how it can be described as statistically significant (line 190). With only four observations, such a regression has very limited inferential power and is highly sensitive to individual years, especially given the large change between 2024 and 2025. The role of Cohen’s d should also be clarified. Cohen’s d quantifies the magnitude of differences between two groups; it is not a statistic for testing or describing a temporal trend. The large values reported for the 2024-2025 comparison therefore show that these two campaigns differ strongly relative to their within-group variability, but they do not provide evidence for a persistent increase through time. Similarly, the significant ANOVA shows that the sampled campaigns differ, but it does not establish the direction of change.
The manuscript should also clarify what was treated as an independent replicate in the ANOVA. If each albedo observation acquired every few seconds along a traverse was entered as an independent data point, the resulting p-values may be strongly inflated. Measurements collected within the same campaign and along the same transects are likely to be spatially and temporally autocorrelated and should not be regarded as independent temporal replicates of a given year. In that case, the nominal sample size would be much larger than the effective number of independent observations, resulting in pseudoreplication. This issue is relevant because albedo is strongly influenced by short-term meteorological conditions and recent snowfall. With only four sampled summers, it seems difficult to distinguish a persistent interannual trend from differences in conditions at the time of each campaign. If the authors want to retain a trend interpretation, the statistical framework needs to be better justified.
I am also not convinced that Section 4.2 represents a true validation of the normalization procedure. MAE, RMSE and correlation quantify how closely the traverse data follow the exponential curve already assumed as the temporal baseline, but they do not independently demonstrate that this curve correctly represents the temporal component to be removed. This creates a sort of circularity: small residuals are interpreted as good normalization, whereas large residuals are attributed to spatial heterogeneity. However, part of those residuals could equally result from an incorrect temporal model. I suggest avoiding the term “validation” unless an independent benchmark is available, and describing this analysis instead as an assessment of model fit or internal consistency. A sensitivity analysis to the decay coefficient would also be useful.
The statement that albedo residuals are “significantly correlated” with slope and TRI (line 215) requires stronger statistical support. No significance levels or confidence intervals are reported, and Table 5 shows that several relationships are weak. With the large number of observations available in some campaigns, statistical significance should also be clearly distinguished from effect size, since even very small correlations may become significant. I also wonder whether ordinary Pearson correlation/least-squares regression is the most appropriate approach given the scatter, possible outliers and heteroscedasticity visible in Figures 6-7. A non-parametric measure such as Spearman’s ρ, and/or a robust slope estimator such as Theil–Sen, could provide a useful robustness check. More importantly, the analysis should account for the likely spatial autocorrelation among consecutive measurements along the same transects.
The interpretation of absorbing impurities remains largely qualitative (this is well stated). The manuscript attributes some of the largest deviations from the temporal decay model to dust, algae and other impurities, and microscopy provides qualitative evidence that such material is present at some locations. However, this does not quantitatively demonstrate that impurities explain the observed albedo anomalies. Since spectral measurements are available, a stronger analysis could examine the spectra associated with the largest residuals and apply established spectral indices or diagnostic approaches for detecting mineral dust and snow algae. Without this type of analysis, the role assigned to impurities remains mainly interpretative.
Overall, I find the dataset potentially interesting, but several of the main conclusions appear stronger than what the current methodological and statistical framework can support. The raised issues affect many fundamental aspects of the manuscript. The dataset could form the basis of an interesting study, but substantial methodological reconsideration and a more conservative interpretation of the results would be required. In its present form, I do not consider the manuscript sufficiently robust for publication in *The Cryosphere*.
More specific comments
Line 27: “causes the rounding of single snow-grains, reduces their specific surface area (SSA) ”
Line 33-36: from what you are stating here, it seems that models are still far from providing comparable results, right? I would stress a bit more that this is currently a limitation, requiring further studies. Now it seems that each model provides different estimates and we are happy about that.
Line 37-40: now it seems that studies carried out on glaciers of Hurd Peninsula are the only one dealing with the high-resolution temporal evolution of snow albedo. I would rephrase to describe that is one of sites where a lot of data are available, but this is not the only site.
Line 49: you talk about “observed diurnal albedo cycles”, but a few lines above you also stated “The natural tendency of albedo is to decay continuously unless reset by fresh snowfall”. So, is this a continuous decline or rather a decline associated with a seasonal cycle? Please explain better this point.
Line60: again it is not clear why you limit the discussion to the South Shetland Islands? Why is this issue so relevant for this area? Of course it is because of the occurrence of ice-free areas, the low-elevation and occurrence of liquid water, which promote the deposition of dust, the spreading of algae and the formation of cryoconite. But now this is not so clear.
Line 91: the error associated with the ELA is the standard deviation calculated considering the whole glaciers in the many seasons considered here? Or is it an estimate from models? I am asking since you use the term “estimated”. Now this is not clear.
Line 99: what findings? More in general, I can’t understand this passage “These findings have direct implications for three areas:”. I can’t get the link and the meaning of the 3-points list here.
Line 114: what kind noise can affect illumination? Is this related to clouds? Instrumental noise, vibration during movement? Please explain a bit the conditions not favorable for this kind of acquisition. It could be useful to reproduce this study in other contexts.
Line 115: the term “body-shadow bias” is not immediately clear, at least to me. I suggest briefly explaining what we are talking about, I guess that is the fact that the operator carrying the portable albedometer partially blocks incoming radiation and therefore reduces the reflected irradiance measured by the downward-facing pyranometer. I would explain better.
Line 117: please expand the acronym: Topographic Ruggedness Index
Line 169: here you talk about “decay experiment”, does it mean that this description is only related to the fixed measurement station? So the acquisition along the transects were conducted differently? Please better explain. The same at line 182, snow grain size was only measured at the fixed site and not along the transects?
Line 203: grey should be light blue, red is missing for the fitting curve
Line 204: you write “The initial snow surface (M1) presented a mean visible albedo (400–700 nm) of ∼0.83”, if I look at Figure 3, initial albedo seems something like 0.530-0.535. Am I missing something?Citation: https://doi.org/10.5194/egusphere-2026-4094-RC2 -
CC2: 'Reply on RC2', Susana del Carmen Fernández Menéndez, 30 Sep 2026
We thank Reviewer 2 for the careful reading of the manuscript and for the detailed comments. We appreciate that the reviewer considers the dataset potentially interesting. We agree that some parts of the manuscript need clarification and reinterpretation.
However, we would like to clarify that the main objective of this work is not to establish a long-term climatic trend from four summers. The main contribution is to analyse, at a spatial and temporal resolution that has not been attempted before in this area, the variability of surface albedo on the glaciers of Hurd Peninsula under clear-sky summer conditions. Acquiring such data is exceptionally challenging, and separating the spatial component from the temporal one—or at least approximating it—is of the utmost importance to calibrate the scarce satellite images that can be found in the area, since cloud cover prevents obtaining images. If we also take into account that the only data to do this would come from pyranometers stationed on the beach without snow or ice, this makes this attempt even more valuable. We have several tracks acquired during different campaigns, but only the tracks obtained under blue-sky conditions were used here. Other tracks were acquired under high clouds or cloudy conditions and, in these cases, the measured albedo is strongly influenced by variable incoming radiation and cannot be interpreted as direct surface reflectance.
We agree that the manuscript should distinguish more clearly between observed patterns, statistical relationships and physical interpretations. We will revise the manuscript accordingly and moderate conclusions that are stronger than the current data can support.
General comment on robustness and suitability for The Cryosphere
Reviewer comment: “The dataset could form the basis of an interesting study, but substantial methodological reconsideration and a more conservative interpretation of the results would be required. In its present form, I do not consider the manuscript sufficiently robust for publication in The Cryosphere.”
Response: We agree that the manuscript requires substantial revision and that some conclusions should be more conservative. In particular, we agree that four independent field campaigns are not sufficient to demonstrate a robust long-term interannual trend. We will therefore remove the expression “statistically significant positive trend” and describe the results as differences among the sampled campaigns, which are associated with contrasting snow and meteorological conditions.
Nevertheless, we respectfully believe that the work is suitable for The Cryosphere after revision. The most important contribution of the manuscript is not the claim of a long-term trend. It is the noveltry and the analysis of spatially distributed albedo measurements at a scale that is usually not available in this area, and the attempt to separate, as far as possible, short-term temporal changes from spatial variability related to snow state, elevation, terrain context and surface impurities. We agree that this separation cannot be complete with the present dataset, and we will make this limitation clearer.
The study is based on a rare in situ dataset from maritime Antarctica, where field measurements of glacier albedo at high spatial resolution are difficult because of logistics, frequent clouds and rapidly changing surface conditions. The dataset combines mobile broadband-albedo transects, a fixed high-frequency experiment, field spectroradiometry, snow-surface microscopy, meteorological observations and terrain analysis.
We will revise the text to clarify that the temporal correction is an empirical reference function derived under specific clear-sky conditions, not a universal law valid for all campaigns. We will also add a sensitivity analysis using different values of and , clarify the treatment of spatial autocorrelation and reduce causal interpretations where the evidence is qualitative.
Introduction
Reviewer comment: The Introduction mixes a general overview of albedo studies with the specific context of the Antarctic Peninsula and Hurd Peninsula.
Response: We agree. We will reorganise the Introduction to separate more clearly: (1) the general controls on snow and ice albedo; (2) the specific conditions of maritime Antarctica and the South Shetland Islands; and (3) the objectives of this study at Hurd Peninsula.
We do not intend to study the whole Antarctic Peninsula or Antarctica in general. This work is focused on maritime Antarctica, especially the South Shetland Islands and, more specifically, on summer albedo variability on the glaciers of Hurd Peninsula.
Field measurement protocol
Reviewer comment: The field protocol should be expanded. It is unclear whether measurements were acquired while moving or at stationary points, and how sensor orientation, height, GPS synchronisation and possible shadow effects were controlled.
Response: We agree that the description of the field protocol in the manuscript is too short. The protocol was described previously in Calleja et al. (2024), where we performed a sensitivity analysis of the pyranometers and glacier measurements. However, we agree that the manuscript should contain enough information to understand and reproduce the acquisition procedure.
The measurements along the glacier were acquired while moving along the transects. The albedometer was mounted on a rigid pole, with one pyranometer facing the sky and the other facing the snow surface. The sensors, datalogger and GPS were synchronised. The instrument was designed as a prototype for these field measurements.
The mobile tracks were processed in QGIS and data acquired under unsuitable conditions were excluded. We selected data measured under clear-sky conditions, close to solar noon, and at an appropriate speed to avoid excessive pole oscillation. Data affected by rapid variations in incoming irradiance were also excluded using the coefficient-of-variation criterion.
We agree that the terms “body-shadow bias” and “illumination noise” should be better explained. We will define them in the Methods section and will provide a more detailed description of the instrument, the sensor mounting, sensor height, acquisition frequency, GPS matching and data filtering. We will also include a full protocol and a schematic figure in the Supplementary Material.
We are currently developing a new version of the instrument, including a GPS, gyroscope and accelerometer integrated in the sensor head. This system is designed to select measurements acquired under optimal levelling and movement conditions, making the post-processing more robust. This instrument is currently described in a manuscript under review. We will not use those unpublished results to support the present manuscript, but we mention this only to explain that we are aware of the limitations of mobile measurements and have worked specifically to improve them.
Temporal normalisation and decay coefficient
Reviewer comment: The exponential coefficient was obtained from one 40-minute experiment in 2024 but was applied to four campaigns. The reviewer also questions the use of P95 as α0.
Response: We understand this concern and agree that the original text was too strong. We do not assume that the decay coefficient is invariant among years, locations or snow conditions. The fixed experiment was performed in 2024 at approximately the mean ELA, where the transition between the accumulation and ablation zones is usually established.
Our original idea was to separate, as far as possible, the temporal component from the spatial component of albedo variability. We measured albedo every second for 40 minutes at a fixed location and then used this empirical function as a reference for the mobile tracks. We agree that this function cannot be assumed to represent all years and all snow states.
For this reason, our intention is not to defend the absolute normalised values as representative of Antarctica or of all Hurd Glacier conditions. The interest is in the spatial patterns and in the relative differences observed along the tracks, after accounting for a first-order temporal effect. We will revise the manuscript to make this point clearer.
We will also add a sensitivity analysis. We have tested different definitions of , including mean, median and maximum values. We will include these tests in the revised manuscript and compare them with the P95 approach. P95 was initially selected because it represents the upper part of the albedo distribution while avoiding isolated extreme values that may be influenced by sensor levelling or specular effects. However, we agree that P95 should not be interpreted as the albedo of the entire traverse at . It will be described as an operational reference value.
We will also test how the residual patterns change when using different plausible values of . The revised manuscript will describe the exponential function as an empirical clear-sky reference function, rather than as a universal decay law.
Metamorphism and solar geometry
Reviewer comment: The observed decay may include both snow metamorphism and changes in illumination geometry. Therefore, it cannot be uniquely attributed to metamorphism.
Response: We agree. This is an important point. The stationary experiment was designed to identify the short-term temporal change in measured albedo under clear-sky conditions, but this change may include several simultaneous processes: snow metamorphism, liquid water evolution, solar geometry and BRDF effects.
The microscopy and spectral measurements show that snow metamorphism occurred during the fixed experiment. We observed grain rounding, grain coalescence and changes in the spectral response. However, we agree that these observations do not quantify the exact fraction of the broadband albedo decay caused by metamorphism alone.
We will revise the manuscript throughout to avoid stating that the complete observed decay is uniquely caused by wet-snow metamorphism. We will instead describe it as a short-term clear-sky temporal signal, observed under one particular surface and illumination condition, with evidence that snow metamorphism was one of the processes occurring during the experiment.
We also agree that residuals after the temporal correction cannot be considered purely spatial. They may include remaining temporal-model mismatch, in addition to real spatial heterogeneity. This limitation will be explicitly stated.
REMA resolution and terrain variables
Reviewer comment: The 8 m REMA DEM and TRI do not represent microtopography at the scale of individual albedo observations.
Response: We agree. We used REMA at 8 m resolution, although higher-resolution REMA products are available for some areas. We agree that the TRI calculated from a 3 × 3-cell neighbourhood represents terrain variability at a scale of several metres to tens of metres. It does not resolve all small features such as individual crevasses, local drainage channels or small snow-surface structures.
We will revise the terminology in the manuscript. We will no longer refer to the DEM-derived TRI as direct “microtopography” at the point scale. Instead, we will describe slope and TRI as terrain-context variables derived at the scale of the DEM.
We will also explain that weak relationships with TRI may result partly from the mismatch between the spatial footprint of radiometric observations and the scale represented by the DEM. We agree that this is a limitation of the analysis.
We will evaluate whether the broad patterns remain similar using a higher-resolution REMA product for the study area. However, the focus of this manuscript remains the spatio-temporal variability of glacier albedo, rather than a detailed study of glacier microtopography.
Inter-campaign differences and statistical interpretation
Reviewer comment: Only four field seasons are available; therefore, the reported interannual trend is not statistically robust. ANOVA may also involve pseudoreplication because consecutive observations along transects are not independent.
Response: We agree with the reviewer. The effective number of independent temporal observations is four campaigns, not the number of albedo observations acquired along the tracks. Therefore, we will remove the claim of a statistically significant long-term trend of +0.016 yr⁻¹.
The revised manuscript will describe the observed differences among the sampled campaigns. In particular, it will state that the 2025 campaign showed higher albedo than the previous campaigns and that this pattern coincided with a season with greater snowfall frequency and precipitation recorded at the available meteorological station. We will not interpret this as evidence of a persistent long-term increase in albedo.
We will also clarify the role of Cohen’s . It will only be used, if retained, as a descriptive measure of the magnitude of the difference between two campaigns. It will not be used as evidence of a temporal trend.
We agree that using all 5-second measurements as independent temporal replicates would produce pseudoreplication. We will analyse spatial autocorrelation along the tracks and estimate an effective sample size or use spatial blocks for the analyses. We will revise the ANOVA-based interpretation accordingly and will avoid using nominal p-values from spatially dependent observations as evidence of interannual change.
Assessment of normalisation
Reviewer comment: Section 4.2 is not a true validation because MAE, RMSE and correlation assess fit to the same assumed exponential curve.
Response: We agree. We will remove the word “validation” from the manuscript. The section will be renamed as an assessment of model fit or internal consistency of the temporal correction.
MAE, RMSE and correlation will be described as measures of how closely the observed transect data fit the chosen reference function. They will not be presented as independent proof that the temporal function is physically correct. We will also clarify that large residuals can result both from spatial heterogeneity and from imperfect representation of the temporal component.
Correlation with slope and TRI
Reviewer comment: The correlations need stronger statistical support, robust methods and consideration of spatial autocorrelation.
Response: We agree. We will report confidence intervals and will complement Pearson correlation with Spearman’s . We will also calculate robust slopes, for example using the Theil–Sen estimator, where appropriate.
We will not focus only on nominal statistical significance, because the number of observations can be large even when the effective sample size is smaller due to spatial autocorrelation. The revised interpretation will focus on the magnitude, direction and consistency of the relationships among campaigns.
We will also make clear that slope and TRI may covary with other factors, including elevation, snow state, surface impurities and illumination conditions. Therefore, correlations do not prove that terrain is the only or direct cause of the residual albedo patterns.
Surface impurities
Reviewer comment: The role of impurities is mainly qualitative and is not quantitatively demonstrated.
Response: We agree. The original manuscript may give the impression that impurities quantitatively explain all major negative residuals, which is stronger than the available evidence.
We have field spectroscopy and microscopy showing the presence of mineral dust, red snow algae and mixed dust–algae assemblages at selected locations. These observations show that the different surface types have distinct spectral behaviour. However, we agree that we do not have systematic co-located measurements of impurity concentration, grain size, liquid water content and bidirectional reflectance along all tracks.
We will revise the text to state that the spectral and microscopic observations are consistent with an influence of impurities in some low-albedo areas, especially where negative residuals coincide with observed dust, cryoconite or red algae. We will avoid claiming a complete quantitative attribution of residuals to impurities.
Where paired spectral observations are available, we will add a more explicit comparison of the spectra associated with clean snow, mineral-rich surfaces and red-algae-dominated snow. Nevertheless, a full quantitative separation of dust, algae and snow-property effects requires additional observations and will be identified as future work.
Specific comments
Line 27: “causes the rounding of single snow-grains, reduces their specific surface area (SSA)”
Response: We will revise the sentence for clarity and accuracy. We will state that snow metamorphism promotes grain rounding and coarsening and generally reduces specific surface area.
Lines 33–36: model uncertainty
Response: We agree. We will emphasise that model-dependent differences in grain-size retrieval remain an important uncertainty and that more field validation is needed.
Lines 37–40: Hurd Peninsula as the only site
Response: We agree. We will revise the sentence. Hurd Peninsula is one of the sites where relevant albedo and glaciological data are available, but it is not the only site where high-resolution albedo evolution has been studied.
Line 49: continuous decline or diurnal cycle?
Response: We agree that the wording is confusing. We will distinguish between the general ageing of snow, which tends to reduce albedo unless fresh snow resets the surface, and the apparent diurnal variation in measured albedo caused by solar geometry, BRDF effects and short-term changes in snow surface state.
Line 60: Why focus on the South Shetland Islands?
Response: We will clarify this point. The South Shetland Islands are particularly relevant because glaciers occur at low elevations in a maritime climate with frequent rain–snow transitions, liquid water during summer, recurrent exposure of ice-free surfaces, deposition of mineral material, and favourable conditions for algae and cryoconite development.
Line 91: ELA uncertainty
Response: We will clarify that the ELA values and their uncertainty are published values from Navarro et al. (2013), calculated over the stated study period. They are not derived from our measurements.
Line 99: “These findings have direct implications…”
Response: We agree. This sentence is misplaced because it refers to findings before the Results section. We will remove it from the Study Area section and incorporate the relevant motivation in the Introduction or Discussion.
Line 114: Illumination noise
Response: We will define illumination noise more clearly. It refers mainly to short-term variability in incoming irradiance caused by thin clouds or changing atmospheric conditions. Movement-induced changes in sensor levelling can also affect measurements. We will explain the filtering procedure used to exclude unsuitable data.
Line 115: Body-shadow bias
Response: We agree that this term needs explanation. We will define it as partial obstruction of incoming or reflected radiation by the operator, snowmobile or support structure, depending on solar geometry. We will also clarify how this potential effect was considered during field acquisition and post-processing.
Line 117: TRI
Response: We will write the full name, Terrain Ruggedness Index, at first use.
Lines 169 and 182: Fixed experiment versus transects
Response: We will clarify this distinction. The decay experiment, spectroradiometry and microscopy were performed at a fixed location. The mobile transect data were acquired separately while moving along the glacier. Snow grain images were obtained only at the fixed site and should therefore be interpreted as evidence of temporal surface evolution at that location, not as direct measurements of grain size along the entire tracks.
Line 203: Figure colour
Response: Thank you. We will correct the colours and ensure that the fitted curve is clearly visible.
Line 204: Visible albedo in Figure 3
Response: We thank the reviewer for noting this apparent inconsistency. Figure 3 shows broadband albedo measured with the paired pyranometers, while the value of approximately 0.83 refers to spectral albedo averaged over the visible wavelength range, 400–700 nm. We will clarify this distinction in the text and figure captions and carefully check the spectral processing and calibration.
We thank Reviewer 2 again for the detailed comments. We believe that the revisions will improve the clarity, robustness and transparency of the manuscript. The revised manuscript will present a more conservative interpretation of the temporal and inter-campaign analyses while preserving the main contribution of the study: a detailed in situ analysis of fine-scale glacier-surface albedo variability in maritime Antarctica
Citation: https://doi.org/10.5194/egusphere-2026-4094-CC2
-
CC2: 'Reply on RC2', Susana del Carmen Fernández Menéndez, 30 Sep 2026
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 157 | 57 | 22 | 236 | 18 | 16 |
- HTML: 157
- PDF: 57
- XML: 22
- Total: 236
- BibTeX: 18
- EndNote: 16
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
## Summary
This study focuses on the acquisition and analysis of in-situ spatio-temporal albedo measurements on an outlet glacier located in a maritime Antarctic setting. The study utlises in-situ measurements acquired during several austral summers between 2018 and 2025. This area sees frequent summer snowfall, which is reflected in the albedo records acquired; snow cover dominates the albedo variability, alongside (lesser) effects attributable to topographic variabiliity and the accumulation of dust and what is identified here to be red snow algae.
While the presentation quality of the manuscript is generally good and includes appropriate referencing, I have a number of major comments which appear to severely limit the potential of the paper to support the conclusions drawn therein. If I have simply misunderstood then I would welcome dialogue on why this is the case and would hope that the manuscript is updated accordingly to address these misunderstandings.
## Major comments
### Insufficient data to support temporal trend analysis.
As the authors note, at this study site there is substantial temporal variability in albedo as a result of frequent summer snowfall. This part of the variability analysis is fine. However, I do have concerns with trying to derive temporal trends from such a small number of temporal sampling points and without explicit consideration in the statistical analysis of the importance of sub-seasonal snowfall to albedo.
In my view if the authors wish to make any assertions about temporal trends in albedo, then different/complementary datasets are required. These could either be from automatic weather stations (although I appreciate that it sounds like these are unavailable), or from remote sensing if there are sufficient cloud-free observations each summer.
### Generalised applicability of the exponential decay function and its use to calculate residuals.
Based on a single measurement period of 40 minutes at a static location, the authors derived an exponential albedo decay function for snowy, clear-sky conditions which they appear to subsequently apply without discrimination to all albedo transects. I have multiple questions concerning this part of the analysis.
1. The authors note that this function was derived under clear-sky conditions. However, no further information is provided, for example day of year/specific date, time of day.
2. Furthermore, what about the surface energy balance at the time these observations were made? The decay of the surface is not only modified by radiative fluxes, it can be modified by other fluxes too; for example the case of weathering crusts in ablation zones (e.g. Schuster, 2001; Stevens et al., 2026).
3. It therefore remains unclear whether it is (always or ever) appropriate to apply this function of modelled albedo decay to each transect time series. insufficient information about the illumination conditions, time of day, day of year of each transect acquisition is provided to be able to assess this.
4. The P95 normalisation (L150 onwards) used to apply this function to each traverse lacks clear motivation. In particular, it is unclear to me why it is valid to take the 95th percentile of the entire traverse series (which is time-varying and so acquired at some time later than time=0) then use that to fix the albedo at time=0. In this respect I would suggest that the reader needs more information on whether there is any dependence of the speed of snow grain metamorphism in relation to the starting broadband albedo. Basically, if the broadband albedo at t=0 were ~0.8 instead of the (very low for snow) 0.54 stated here, then would the decay function found here still stand?
4. It's very unclear to me what insights the residuals plotted in Fig. 5 (right hand column) reveal, given that it isn't clear that applying the exponential model is appropriate.
5. The decay function is invoked as dependency for multiple downstream analyses, including for example L221 where it is stated that 'steeper and rougher terrain yielded albedo consistently below the model prediction in both years', and with its use to derive residuals to correlate against (Figs. 6 and 7). Yet it is completely unclear why this 'model prediction' is appropriate for thie use case.
6. To fully support the discussion and conclusions of Sect. 5.1 I would find it important to further reflect on representativeness of this single measurement set acquired on a single day. For example, is it typical of a normal day on Livingston Island, or were the specific energy balance conditions very rare for this area?
7. At L321-322, I do not understand why this 'residual analysis' permits to identify impurity signals?
### Identification of impurities
I have two key areas to follow up on here. First, as far as I understand, the identification was effectively done only visually, i.e. without follow-up laboratory analysis. It's unclear to me that this is adequate, particularly given the very limited pieces of 'photographic' evidence presented here rather than, for example, summary statistics of a dataset with sufficient statistical power. There is no detailed information provided about sampling protocols, number of samples, replicates, locations/times and so on. This information is critical to understand the reliability of these claims.
There are also specific issues with the analysis presented which revolve around a lack of clarity concerning the underlying surface type. For example, L330 talks of both snow and (cryoconite-contaminated) ice and refers to Fig. 9, yet Fig. 9 is labelled as 'snow surface impurities' for all three surface types. Yet it is essential to differentiate between snow and ice surfaces given their major structural differences. See an abundance of literature on the subject. Furthermore, this is important from the perspective of accurately identifying the algal species: as far as I'm aware, snow algae (i.e. Chlamydomons nivalis) does not live on bare ice surfaces, which are rather colonised by Ancylonema nordenskioldii and A. alaskanum.
## Minor comments
Graphical abstract: In the panel 'Observed albedo vs decay model', it isn't clear that what is being shown here is some kind of space-time substitution on the x-axis, i.e. the x-axis actually is time for the decay function but implicitly space for the observations. This seems problematic, especially for the graphical abstract.
Methods: How was the albedometer kept level on the traverses? Presumably it was not mounted in a gyroscope. Please provide an indication of the error associated with not keeping the albedometer levelled. (Though I note that a shadowing correction was applied)
Figure 1a: No information on background shading/image is provided.
Fig. 1 / Fig 5: What is the direction of the snowmobile tracks? This is important so that they can be interpreted along the time-elapsed axis of Fig. 5.
L119-124: contains some repetition of L109-111.
I struggled to understand the systemically lower albedo found at higher altitudes (>~270 m). Please reflect further on this analysis.
Fig 5: The last sentence of the caption contains a description which would be better placed in the Results.
L215-225: Too much repetition of statistics in the text that are found immediately below in Table 5 - it is likely okay simply to refer to the table instead of quoting the stats in the text.
Fig 8: Second sentence of caption has missing period. Third sentence to end of caption belongs in the Results text, not in the caption.
L276-279: this section seems confused. There are multiple references to the elevation zone 'below 200 masl' but also to the accumulation zone, which surely is ~the area above 200 m asl?
L280/281: Is there any reason preventing remote sensing measurements being used to accomplish the same task?
Figure A1, Table A1: I did not find any references to these appendixes in the main text.
Data availability: only the 2025 data are available at the link supplied.
Code availability: followed the github link provided. No public repositories are available at this link.
Statement on inclusion in global research: No 'in-country partners' are listed anywhere in the research. This statement feels like it was not written for this study.
## References
Schuster, Corinne Joanne, "Weathering crust processes on melting glacier ice (Alberta, Canada)" (2001). Theses and Dissertations (Comprehensive). 489.
https://scholars.wlu.ca/etd/489
Stevens IT, Cook JM, Chevrollier L-A, et al. The formation and evolution of the supraglacial weathering crust on the Greenland ice sheet. Journal of Glaciology. 2026;72:e30. doi:10.1017/jog.2026.10146