Melt probabilities and surface temperature trends on the Greenland ice sheet using a Gaussian mixture model

Clarkson, Daniel; Eastoe, Emma; Leeson, Amber

doi:https://doi.org/10.5194/tc-16-1597-2022

Articles | Volume 16, issue 5

The Cryosphere, 16, 1597–1607, 2022

https://doi.org/10.5194/tc-16-1597-2022

© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

The Cryosphere, 16, 1597–1607, 2022

https://doi.org/10.5194/tc-16-1597-2022

© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 16, issue 5

Research article

04 May 2022

Research article | 04 May 2022

Melt probabilities and surface temperature trends on the Greenland ice sheet using a Gaussian mixture model

Daniel Clarkson et al.

Download

Final revised paper (published on 04 May 2022)
Preprint (discussion started on 30 Aug 2021)

Interactive discussion

Status: closed

RC1:
'Comment on tc-2021-259', Anonymous Referee #1, 16 Sep 2021

General comments

I think that the authors have done a credible job and that their results are solid. But I also wonder if they have not overcomplicated the analysis. Why not just use the ISTs and IST trends instead of the mixture model? What value added information is provided by the model? I suspect that there is a good reason but it needs to be explained. Please state the reason succinctly in the Abstract, Introduction and Conclusion. Also there seems to be some confusion with the words: pixel, location and points. If I am correct, then they all refer to 'cells' in the gridded dataset. Please be consistent or describe the differences.

Specific comments

Line #

22        Instead of the word ‘exceeding,’ I would say ‘equal to or exceeding’

23        you should reference the paper that describes this dataset instead of the reference shown – see below (Hall et al., 2018)

24        this is incorrect; see below for the corrected citation for Wan and Dozier (1989)

27        GCMs should read GCM

28        please add “under clear-sky conditions” after the word ‘measurements’

33        I would delete ‘and future events’

38        these are not actually pixels but they are values gridded into cells; therefore, on this line and elsewhere in the manuscript, instead of writing ‘pixels’ you should write ‘cells’

50        daily should read ‘near daily’

50        ‘the dataset’ should probably read ‘this dataset’

51        please check the time span of the dataset; I believe it ends on 12/31/2020 and not in 2019

54        I would replace the word ‘points’ with ‘cells’

58        delete the words ‘a slight’

61        add ‘because of cloud cover’ after ‘summer months’

Fig 1 caption: replace the words ‘non melting’ with ‘available;’ instead of ‘locations,’ say ‘cells’

72        delete single parenthesis

99        please check original reference; this should depend on percent salinity

121      there is also the +/- 1 deg C accuracy to consider

130      is ‘sample’ the same as ‘point,’ ‘pixel’ or cell? If yes, please be consistent and use ‘cell.’

134      do you mean greater than or equal to -1 deg C?

Figure 4 in (a) do you mean greater than or equal to -1 deg C? The heading should say ‘observed.’

138      since you are referring to the glacier facies/zones, you should cite Benson (1962).

151, 159, 172 & 173 and elsewhere   are ‘locations’ the same as cells? If so, then please use ‘cells.’

192-193           is this a statistically significant difference?

203      please delete the word ‘etc’

206      the spatial resolution of the product is 0.78 X 0.78 km, not 1 km as stated

206      after the word ‘resolution,’ please add the following “(cloud-cover permitting)”

Additional or corrected references:

Benson, C.S., 1962: Stratigraphic Studies in the Snow and Firn of the Greenland Ice Sheet, US Army Snow Ice and Permafrost Research Establishment, CRREL Research Report 70.

Hall, D.K., Cullather, R.I., DiGirolamo, N.E., Comiso, J.C., Medley, B.C. and Nowicki, S.M., 2018, A multilayer surface temperature, surface albedo, and water vapor product of Greenland from MODIS, 10(4):555.

Wan, Z. and J. Dozier, 1989: Land-surface temperature measurement from space: physical principles and inverse modeling, IEEE Trans. Geosci. Remote Sensing, 27(3):268-278.

Citation: https://doi.org/10.5194/tc-2021-259-RC1
- AC1: 'Reply on RC1', Daniel Clarkson, 23 Nov 2021
  
  The authors would like to thank the reviewer for their consideration of the paper and for their comments and feedback. We are currently working on a revised version of the paper to address the comments and suggestions made. The comment on the need for a more detailed explanation of the value for the model compared to analysis of only the observations is reasonable and will be focused on in the revised version. We will amend the abstract, introduction, and conclusion to include a clearer explanation of the reasoning behind the modelling approach in addition to the line-specific comments regarding spelling, grammar, and references. The revised version will include a more detailed response to the above recommendation and the recommended corrections. Thank you for all of the comments made and for the time spent reviewing our work.
  
  Citation: https://doi.org/10.5194/tc-2021-259-AC1
RC2:
'Comment on tc-2021-259', Anonymous Referee #2, 11 Oct 2021

This short paper presents an innovative method to derive melt extent from MODIS surface temperature (ST) based on its statistical distribution. While the presented work seems to be statistically robust, I don’t see the interest of this complex method with respect to a simple one based on TS > 0°C. If TS values larger than 0°C are sometimes retrieved in MODIS, it is due to the presence of rocks or impurities at the surface of the ice sheet (ST > 0°C can also been seens in the PROMICE AWSs data set). Therefore a comparison with a melt extent independent data set (like a microwave product) is needed to show the interest of this new method. Moreover, to build the statistic model, several years of observations are needed by assuming that the surface conditions remain similar. But what occurs, if rock or impurities appear in surface after a summer with a huge melt? Are the algorithm enough robust to deal with such problem? Idem, what about percolation zone pixels becoming bare ice zone pixels through the recent summers? I suspect that the statistic distribution of these pixels could change... More details about these potential issues are also needed for me.

Citation: https://doi.org/10.5194/tc-2021-259-RC2
- AC2: 'Reply on RC2', Daniel Clarkson, 23 Nov 2021
  
  The authors would like to thank the reviewer for their consideration of the paper and for their comments and feedback. We are currently working on a revised version of the paper to address the comments and suggestions made. We aim to give a much clearer motivation for the model and the reasons to use the model compared to using solely the observations and the differences in approach to using a melt extent data set. The comments regarding potential changes in the statistical distribution of certain pixels are valid and more work will be done to provide a more detailed explanation of the model's assumptions and why they have been made. A more detailed response to the comments made will accompany the revised version of the paper with full explanations as to the changes made. Thank you for all of the comments made and for the time spent reviewing our work.
  
  Citation: https://doi.org/10.5194/tc-2021-259-AC2

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision

ED: Reconsider after major revisions (further review by editor and referees) (23 Nov 2021) by Stef Lhermitte

AR by Daniel Clarkson on behalf of the Authors (11 Jan 2022) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (12 Jan 2022) by Stef Lhermitte

RR by Anonymous Referee #1 (14 Jan 2022)

RR by Anonymous Referee #2 (21 Jan 2022)

Suggestions for revision or reasons for rejection

Thanks for this revised version and your responses. But I'm continuing to think that a comparison with a microwave based melt product is needed, although I agree that your product is well more than a simple binary melt extent data set. It is nice to show and discuss melt and temperature evolution but we don't know how your product performs. Is it reliable and robust? Is it able to detect the extreme melt events (mostly driven by clouds) knowing that only the cloud-free days are considered here? In the accumulation zone, melt occurs when there are clouds and in the ablation zone, only ~ 50% of days are cloud free.

Therefore, you need to show, at least over one season (e.g. summer 2019) what is the interest and robustness of your complex model with respect to a microwave data set (Can your model be applied to the microwave brightness temperatures?). An other validation, if your prefer, would be to compare the melt at the PROMICE AWS's using measured temperatures as the "true" melt data set.

As cloudiness has been changing and is different between the 2 decades 2001-2009 and 2010-2019, can these 2 decades be compared in Section 3.3 if you consider only the cloud-free days? How do your temperature statistics compare with the ones using surface temperate from ERA5 for example ?

Finally, with respect to Casey et al. (2017), how could your product be impacted by the changes of sensors in the MODIS data set knowing that your model needs no change in the surface conditions?

Ref: Casey, K. A., Polashenski, C. M., Chen, J., and Tedesco, M.: Impact of MODIS sensor calibration updates on Greenland Ice Sheet surface reflectance and albedo trends, The Cryosphere, 11, 1781–1795, https://doi.org/10.5194/tc-11-1781-2017, 2017.

Hide

ED: Reconsider after major revisions (further review by editor and referees) (24 Jan 2022) by Stef Lhermitte

AR by Daniel Clarkson on behalf of the Authors (14 Mar 2022) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (15 Mar 2022) by Stef Lhermitte

RR by Anonymous Referee #2 (24 Mar 2022)

ED: Publish as is (28 Mar 2022) by Stef Lhermitte

AR by Daniel Clarkson on behalf of the Authors (11 Apr 2022) Author's response Manuscript

Short summary

The Greenland ice sheet has seen large amounts of melt in recent years, and accurately modelling temperatures is vital to understand how much of the ice sheet is melting. We estimate the probability of melt from ice surface temperature data to identify which areas of the ice sheet have experienced melt and estimate temperature quantiles. Our results suggest that for large areas of the ice sheet, melt has become more likely over the past 2 decades and high temperatures are also becoming warmer.