Temporal Variability of Surface Reflectance Supersedes Spatial Resolution in Defining Greenland’s Bare-Ice Albedo

Round 1

Reviewer 1 Report

Bare-ice albedo keeps poorly characterized and parameterized in numerical models. This study investigates temporal variability in the reflectance/albedo of the bare-ice using a multi-scale analysis on finer-resolution site albedo observations and Sentinel-2 reflectance products and the coarse resolution MODIS reflectance/albedo products. A series of parameters in the ice ablation equation are calculated to simulate bare-ice melt. It is concluded that regional melt model performance could be improved through involving temporal dynamics of bare-ice albedo, and inappropriate bare-ice albedo schemes yield large errors in seasonal ice ablation.

One main confusion is about the albedo/reflectance spatial pattern analysis (one of principal part of the content), while none of 2-D spatial distribution maps are shown in the manuscript. Some comments are itemized as follows.

1. Line 318-320, “Similarly, the significant correlations between MOD10A, S6 reflectance, MOD09GQ and Sentinel-2 (0.33 < r < 0.93, p < 0.01) … the locality (Figures 2c-e)”. Here add each correlation coefficient beside the targeted albedo data source. Also, I am confused about the correlations. From this sentence, I understand that the correlation coefficient is between MOD10A1 and Sentinel-2, between S6 reflectance and Sentinel-2, between MOD09GQ and Sentinel-2. However, to calculate the correlation coefficient between satellite-derived albedo/reflectance and site observed albedo is meaningful.

2. Line 327-329, “acknowledging the potential influence of cloud conditions on surface reflectance [23], DPP3 …SNBW …were consistent with similar records from clear-sky days.”. Cloudy-sky reflectance is similar to clear-sky reflectance, why? More explanations from other aspects should be added here, now that the author acknowledges the influence of cloud on surface reflectance. Otherwise, such this statement is incomplete.

3. Line 394, the author mentions variations in the spatio-temporal patterns of surface reflectance. The previous analysis focuses on the temporal variations of different albedo/reflectance products, nothing about spatial distribution pattern. Different spatial resolutions or regional mean values are linked to the sample problems. These could not represent the spatial distribution of reflectance. In order to conclude spatial pattern, it is better to show the spatial distribution of different albedo/reflectance products in ROI.

4. Line 425-432, repeat sentences. Please delete one sentence. Also. The parameter A and B is unclear. More description about A and B is needed.

5. [3.2.1 Spatial Variability], I expected to see figures about the 2-D spatial distribution of reflectance in DOI. However, only 1-D frequency and melt changes with reflectance/albedo. If none of 2-D spatial distribution is added, then the title should be changed to a more appropriate one.

6. MOD10A -> MOD10A1. Please check the whole manuscript.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Review of ”temporal variability of surface reflectance supersedes spatial resolution in defining Greenland’s bare-ice albedo”

 

In this manuscript, the authors detail an effort to investigate the magnitude and spatiotemporal variability of surface albedo and reflectance quantities of a bare-ice region located in the ablation region of western Greenland. In situ, UAV and satellite measurements are used as data sources, with the end objective being to assess whether accurate characterization of temporal variability or spatial resolution of the bare-ice zone is more important to estimating the overall ice ablation amount.

Overall, the problem being addressed is one that certainly needs addressing, and the manuscript is topically fully suited to RS. Both the fieldwork and the data analysis have commendable breadth and scope. However, some aspects of the description did leave me somewhat confused, meaning that a revision of the manuscript is still needed to sharpen and clarify the content before publication is considered.

My main comments are as follows:

1. A bit more care needs to be taken with the various uses of albedo and surface reflectance terms and data. As I understand it, only MOD10A1 and the S6 measurements here are estimates of ice albedo, others are processed into nadir-viewing surface reflectance. Of these, several specific issues arise:
   1. it should first be noted that MOD10A1 is a black-sky albedo estimate, i.e. valid under unidirectional illumination – which fortunately mostly seems to fit with the clear-sky conditions of 2016, but this still needs to be explicitly noted, as diffuse illumination is never equal to zero in real-world conditions. Also please specify that MOD10A1 is a shortwave broadband albedo estimate.
   2. Also, MOD10A1 contains to my knowledge an anisotropy correction which assumes snow – not bare ice – cover. What effect does this have on the analysis over the bare-ice ROI?
   3. S-2 was processed into nadir-viewing surface reflectance, yet very little description is given in section 2.2.1 to e.g. the atmospheric correction – what aerosol loading did you assume, what water vapour or ozone column? I principally agree that settling for surface reflectance analysis is better than attempting to enforce some, likely invalid, surface BRDF model, but the processing description needs to be more specific.
2. You compare S-2 with the spectrometer and UAV measurements, implying that S-2 surface reflectance is valid for the 400-700 nm waveband to be consistent with SNBW. Which S-2 bands were used as input, therefore, and how did you combine them?
3. Ablation model: I am not familiar with the ablation model used, so it seemed surprising that the longwave energy component is tied only to 2m air temperature. I guess that a reasonable correlation between cloud cover and 2m Ta would exist over the ice sheet margins, but is the relationship really robust over the whole summer – and any summer? Or is there a chance that the ablation comparison in Fig 4c is robust by happy accident given the largely cloud-free conditions of summer of 2016?
4. Table 1 vs. Figure 4: As written, I read Table 1 so that the fixed bare-ice albedo shows the best metrics with highest NSE and d, and lowest MAE/RMSE. Yet all of the text in section 3.3.2 makes no mention of that? Both the table and text cannot be correct, please check them.
5. Please improve the resolution of all result figures, in particular 4 and 5. In a printout, the dashed grey and all yellow lines in Fig 5 faded to near-invisibility.
6. On lines 474-480, you state that the melt underestimation during early-mid investigation period may be related to ice algae. If so, then the implication would seem to be that the algae contribute very little to the observable MODIS surface reflectance, as none of the MOD10A1-based albedo estimates produced appropriate ablation estimates? And given that the PDD-related albedo estimates fared no better, is the implication that PDD-related melt estimation methods also fail to capture any signal from algae blooms? That would be somewhat surprising, given that one could well assume that 2m air temperature would at least partially correlate with algae growth?

Specific comments per line:

142-143: Why use a 24-window and absolute values of reflected radiation? Reflected radiation cannot be negative in any case.

248: 600 x 600 grid cells covered which area? The UAV-covered area shown in Fig 1?

309: use of word “periodic” suggests to me that the changes should be somehow regularly occurring. Fig 2c does not suggest regularity, so please revise. Also, why is the y-label in 2c labeled “reflectance”? AWS measures with pyranometer pair, so it’s an albedo measurement, correct?

Fig 3: Why does S-2 exhibit the largest distribution changes between overpasses, despite it having the coarsest spatial resolution and therefore the “smoothest” surface?

459-460: Please briefly explain the Nash-Sutcliffe criterion and Willmott Index, these are not everyday terms.

454: broken sentence, please complete.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The reviewer is really very grateful to the authors for such an outstanding research and article on so vital topic of definition of Greenland's bare-ice albedo and about its temporal variability of surface reflectance superseding spatial resolution in its definition. The authors demonstrate considerable spatial and temporal variability in the reflectance of the bare-ice environment using a multi-scale analysis that draws together and compares finer-resolution ground-based spectrometry, UAV- and Sentinel-2 derived reflectance observations and the coarse resolution MODIS reflectance and albedo products. 
Authors’ analysis highlights daily-to-seasonal synoptic variability in bare-ice reflectance, evident in both the spatially averaged reflectance for their 1300 m2 study area along with the actual reflectance distributions measured in-situ with a field-spectrometer. Modeling bare-ice melt with these reflectance distributions yields no significant difference compared to a Gaussian distribution, but highlights the importance of accurately describing the areal mean albedo. Further assessment exploring the temporal variability in bare-ice albedo emphasizes the critical importance of parameterizations at daily-to-synoptic resolution, with inappropriate schemes yielding large errors in seasonal ablation of up to 11%. The authors argue that future research efforts must provide robust albedo evaluations for heterogeneous bare-ice across the range of available observational platform scales, and further refine evolving bare-ice albedo parameterizations using temporally and physically based descriptors.
So the authors propose that albedo parameterizations can be improved by (i) quantitative assessment of the representativeness of time-averaged reflectance data products, and, (ii) using temporally-resolved functions to describe the variability in impurity distribution at daily time-scales. They conclude that regional melt model performance may not be optimally improved by increased spatial resolution and incorporation of sub-pixel heterogeneity, but instead, should focus on the temporal dynamics of bare-ice albedo.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

There are still some points needed to be illustrated.

1. Line 338-340, Cloudy-sky reflectance is similar to clear-sky reflectance, why? The authors have not added any explanation after the last revision. More explanations from other aspects are needed here, now that the author acknowledges the influence of cloud on surface reflectance.
2. Line 432-436, “To simulate these changes…behavior of percentage variables (e.g., albedo) [85]”. The authors mention beta probability distribution to be well-suited to modeling the random behavior of percentage variables such as albedo. I am confused about this statement. Albedo, an inherent property of objects, is a percentage variable. It is affected by snow-related variables, sky conditions, solar zenith angle, terrain and surface pollutant deposition. It is not the random. The authors should correct these sentences.
3. In addition, it is better to give the equation of beta distribution function, so as to clearly know the roles of parameter A and B.
4. Line 436-437, “Here, again using DOY204 as the ‘control’, we defined parameters A and B”. I am not familiar with how to calculate parameter A and B. One or two sentences to illustrate the calculation process of A and B helps to improve the understanding of the two parameters. Also, the reasons why use DOY204 as the control should be involved here.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The authors have satisfactorily addressed my comments and corrected the erroneous result table, improving the robustness of their arguments. I see no further obstacles to publication.

Author Response

We thank the Reviewer for the positive feedback on our first set of revisions, and are pleased the Reviewer felt no further changes were required.