Warming Has Accelerated the Melting of Glaciers on the Tibetan Plateau, but the Debris-Covered Glaciers Are Rapidly Expanding

Round 1

Reviewer 1 Report

In this study, the authors produced long-term clean and debris-covered glacier areas using high-resolution landsat images, carefully analyzed the results and concluded interesting and important scientific results. The results were consistent with my personal knowledge of the cryopshere change in this area. The paper was well written. Most places were appropriate. I have only a few suggestions as follows.

 

Major:

1.Section 4.5,compared to the annual precipitation, it is suggested to use annual snowfall, or precipitation in winter months at least, for analysis.

 

Minor：

1.line 328, and glacier thickness

 

Grammar:

1.line 63. combat

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

 

Hu et al. construct an inventory of clean ice and debris-covered glaciers in the Tibetan Plateau using extensive amounts of Landsat scenes and a double random forest classification algorithm. The breadth of Landsat data and the geographic scope of the inventory are impressive. They present their methods clearly and, in doing so, compile a valuable dataset. However, how they analyze their dataset has shortcomings. Based on the analyses presented, I am not yet certain that their conclusions are supported by the data.

 

Comments:

 

Line 45 - 51: Separate this sentence into multiple sentences. I’d recommend one sentence for each subregion.

 

Line 54: The phrase ‘mountain movements ‘is unclear throughout the document. Do you mean landslides adding debris to the surface of the glaciers?

 

Section 3: Your Methods section is very clear and well written. Also, its figure (Figure 2) is very informative!

 

Line 188 – 199 (and Figure 5): Further analyses are needed -- in particular, in regard to trends in the area change rates. I enumerate some points of concern in the following four comments.

 

Figure 5a: a bar chart with different time bins is misleading and obscures true temporal trends. I recommend remaking the future as a line graph with time (and not time bin) as the x-axis.

 

Figure 5b – Figure 5d: With your time-series plots, making the x-axis time (instead of bin number) will allow you to find a meaningful trendline for the data.

 

 Figure 5b – Figure 5d: For determining if the change rates are changing with time, fitting models to the area data seems more appropriate than fitting models to the area change rate data. I suspect that your change rate data on the y-axes of Figure 5b – 5d are calculated by differencing the areas between two-time bins. If that is the case, then your approach is highly suspectable noise in the data. Instead, if you fit models to the area data themselves, then you can find a truer trend in the data. If you find that a linear regression of the area vs. time plots outperforms a model with curvature (e.g., quadratic curve of best fit), then your claim about accelerating advance/retreat rates would not be supported by the data. If a non-linear model fits the area vs. time data better than a linear regression model, then your data support your claims about changing rates of advance/retreat. To determine which model performs better, make sure to use a metric like the Akaike Information Criterion (AIC) or adjusted R^2 value instead of just the ‘standard’ R^2 value.

 

Figure 5c & 5d: It seems your trends -- accelerating loss for clean ice and accelerating growth for debris-covered glaciers – may be strongly influenced by noise. The area change fluctuations between the last three data points are significantly larger than trends in the data. First, have you performed statistical tests to determine the significance of the trends? Or was your conclusion based just on the slope of the trendline? Make sure to include in the text whether or not the trendlines are significant. Furthermore, even if you find statistically significant trends, discussion in the text is warranted about the magnitude of variability and its influence on the trends you find.

 

Figure 6: How were the trends calculated for each region? Did you perform a linear regression on all of the data to find a retreat (or growth) rate? Or did you instead just subtract the last area from the first? I feel the first approach would be the better one. If your trends come by just differencing two points, then noise in those points could significantly influence the value of the trends. Regardless of which approach you used, make sure to discuss explicitly how you found the trends in Figure 6.

 

Line 216 – 217: What test did you use for determining that the area-elevation curves are normally distributed?

 

Line 222 – 226: Explain where and how there has been glacier growth at higher elevations. I suspect you are seeing snowfall being classified as glacier extent in some of the later dates, which you have interpreted as glacier advance at higher elevations. However, it is possible there are glacier advances at higher elevations. If so, then illustrate that point by finding some locations where you see glacier advance at higher elevations and can be differentiated from ephemeral snow cover.

 

Figure 7b: The twin y-axis plot should be remade as a single y-axis plot. Right now, your left axis has a zero point while the right y-axis is all positive values. This design is misleading as it appears that both types of glaciers (clean and debris covered) as losing more elevation at lower elevations and gaining more at higher elevations. In reality, your data suggest that debris-cover glaciers are gaining at all elevations but slightly more at higher elevations. However, it is worth considering if a change from 0.62% gain at the lowest elevation to 0.63% at higher elevations is distinguishable from uncertainty. If you plotted both the orange and blue curves on the same y-axis, I think the elevation trends for debris cover glaciers (orange) would look like a flat line. That visual may suggest a reinterpretation of elevation changes for debris-covered glaciers.

 

Figure 7b follow up: Something is amiss with the numbers on both y-axes.

 

Line 230 - 231: I think a word is missing. You say that ‘the slope was too steep.’ But for what is the too-steep value? Greater than 50 degrees? If so, make that threshold explicit.

 

Line 233: I think there is a typo regarding which class of glaciers is retreating the fastest in the 10 to 40 degree range. I think you mean the clean-ice glaciers are retreating fastest in this range. Debris-covered glaciers seem to be advancing at all slopes.

 

Line 234: The phrase ‘the expansion of debris-covered glaciers accelerated with decreasing slope’ does not seem accurate. The growth rate is constant for the points between 0 degrees and 40 degrees and slightly smaller at 50 degrees (but likely within uncertainty). The only deviation from a constant retreat rate is in the 60-degree slope bin. However, there is not much glacial area in that bin, meaning I would be hesitant to draw conclusions from that outlier.

 

Line 241: ‘change rate of debris-covered glaciers’ (instead of just covered glaciers)

 

Figure 9b: Remake the one rose diagram into two diagrams. Having positive and negative radius values in one diagram makes the figure hard to interpret. Instead, make one diagram for clean ice glaciers, where the radius is the retreat rate, and another diagram for debris-covered glaciers, where the radius is the growth rate.

 

Section 4.5 and Figure 10: I see only one temperature and precipitation plot over the entire region. Did you also undertake a similar analysis per subregion? If so, how homogenous are the temperature and precipitation signals? To draw climatic conclusions for the entire Tibetan Plateau, there should be more than one record of temperature and precipitation – unless you can show that the curves in Figure 10 are representative of the entire region.

 

Figure 10 follow up (addition to comment for Line 222 – 226): There seems to be an upward trend in precipitation in the last three data points. That increase in precipitation may explain why you classified more glacial area at higher elevations in more recent years. With more precipitation, there would be more snowfall. However, that extra classified area is likely not glacial extent. Instead, it is likely snowfall that is being falsely classified as part of a glacier.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper, based on the Landsat series of images study clean and debris-covered glaciers on the Tibetan Plateau, provide the change data and process information of the two glacier types from 1985 to 2020. The conclusion that climate warming was a main driver of glacier change is of very important remark. The research provide scientific basis for management of glaciers affected regions on the Tibetan Plateau in the context of climate warming.

Some Comments and suggestions

1  Abstract

The abstract has too many words and needs to be compressed

2  Fig 1

TB legend, basin ? or should be “sub-area”; Major rivers need to be represented by prominent color and line.

3  The article uses “basin”, maybe it is better to use “subarea”

4  How do two types of glacier changes in the southern margin of the Qinghai Tibet Plateau (Pamirs, Karakoram, Himalayas) affect the Indian monsoon on the plateau environment. 

5  It is better to further discuss the climatic type and boundary of glacier development in the Qinghai Tibet Plateau

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

Review Report: Warming has accelerated the melting of glaciers on the Tibetan Plateau, but the debris-covered glaciers are rapidly expanding

 

 

Dear Authors,

This manuscript is generating inventories of the glaciers on the Tibetan Plateau based on the Landsat archive and uses the Random Forest machine learning algorithm. Inventories are generated for several time periods and the changes in glaciers are evaluated. I have some concerns with the way the paper is presented and some substantial changes are necessary to justify a publication in an international journal. I encourage the authors to clarify some aspects:

 

·       Title: Why Tibetan Plateau and not Qinghai-Tibet Plateau as appear in most publications?

 

·       In the abstract, the methodology is omitted.

 

·       The manuscript adds some confusion between moraines and debris-covered glaciers. Please clarify throughout the manuscript. There is not a clear presentation of the difference between clean and debris covered glaciers from the morphological and remote sensing points of view.

 

·       Many long sentences that are hard to follow.

 

·       There is a subchapter of the study region but only describes the location. It needs elaboration with a proper description of the study area.

 

·       The methodological steps are unclear and weakly described with many unclarities. It is hard to reproduce the work based on this description. The methodology needs deep revision.

 

·       Why the accuracy of the second classification is lower than the first one? Shouldn’t be better? Why haven’t been tried other machine learning algorithms and was chosen Random Forest only based on literature?

 

·       For the climate why not use the data from stations than the reanalysis products that have some limitations? Or why not both to be compared?

 

·       Discussions are insufficiently written. Are repeating the results and only a few comparisons to other studies but for different time periods that cause the differences. Nothing concrete in the reference to climate change. It requests to be rewritten to better emphasize the importance of these results, technical constraints, proper comparisons to other studies, and areas. Not clearly discussed why the differences between the clean and cover-debris glaciers regarding their changes, location, and environmental controls.

 

·       The conclusions are repeating the abstract, while it needs to be complementary.

 

Table 1: What is the empty image rate and how is calculated?

Table 4: Not clear what the notes are referring to. Does it refer to an inventory of the glaciers from all over China? Please add references. How are compared to the ones from the Tibet Plateau if include glaciers from other regions?

Figure 1: The elevation scale suggests that all the dark blue surfaces (e.g. India) are at an elevation of 6 m, please revise. The map seems elongated, please check the projection.

Figure 3: Specify in captions what is referring by “(2)”.

Figure 5: Axes labels are too small to be read.

Figure 6: Please increase the size of the labels on the map. If all basins are labeled the colors are not necessary anymore. Could be removed to better see the glacier inventory.

Figures 7-9: Please remove the border of the legend and the minor gridlines from the plot area. Font size could be increased.

 

 

Apart from these unclear aspects, I have some specific comments:

Line(s) in manuscript : Comment

 

Abstract

1.18: “mountain movements” sounds a bit strange. Please change it or elaborate. Please change “moraine cover” to “debris cover”

1.22: Please change to “…and temporal changes were analyzed under the background of climate change”

l.22-24: Not clear if here is referring to both types of glaciers.

l.25: Change to “…..an expanding rate of 0.62 %/a.”

l.28: Unclear as written what is referring to the trend.

l.29: Please start with a new sentence.

l.35: What do you refer to by scientific management?

 

Introduction

1.44-48: Please add some values/examples.

1.44-51: A very long sentence, please split it into shorter ones.

1.54: Please elaborate on what is referring to mountain movements.

l.55: Change to “debris cover”

l.56-59: Long sentence, please rephrase.

l.58: thermal properties?

l.68-72: Hard to follow, please rephrase.

l.79-83: Hard to read, please keep it short.

 

Study region and data

1.86: Coordinates are not necessary, remove them to keep this sentence short.

1.87: What is the Alpine Range?

l.96: Please describe this fill function and add references.

 

Method

1.112-114: A bit unclear this procedure, please clarify.

l.113: 60 points or 60%?

l.120: Maybe proper is “merged” than synthesized unless it is a specific operation that needs to be described then.

l.121-123: Unclear as written, please rephrase.

l.125: Not clear what minimum value it is about.

l.131: Change to “texture and topographic features”.

l.136-137: References requested or explanations for some terms such as angle second moment or inverse different moment.

l.139: As written could be understood that only topographic features are 25 and not all. Rephrase and could move this information to the beginning of the listing.

l.147-148: As written suggest that were used three categories of glaciers and four categories of snow. Please rephrase.

l.163-164: Redundant information.

l.169: Please clarify what represents these “pixel patches”.

 

Results

l.176: Repeating information.

l.248-256: This paragraph is more suitable for methods than for results.

l.265-266: “….for times that of precipitation”. Greater?

 

Discussion

l.272: This paragraph is not suitable for discussions, maybe at methods, or if remain at discussion need to be more elaborated and compared to other studies.

l.276: It is not %?

l.277-279: Unclear as written, please clarify.

l.280: Empty pixel rate of 1% is referring to the cloud coverage of 1%? Please clarify this.

l.290-291: What are the “existing research results”? From other studies? Is referring to an inventory? Add reference.

l.293: “2000 [51]”

l.293-295: That sounds like adding a lot of errors and overestimations in the glacier delineation. Could be really compared to that?

l.295-296: Can’t see the point here.

l.308-314: How can be compared if the time periods are different? Of course, there will be differences.

l.322: This short paragraph only states that the temperature and precipitation are not enough to link the glacier change to climate change. Thus, it is stated that the study didn’t add anything now.

 

Conclusions

l.339: The sentence looks unfinished.

l.357-361: But nowhere was explained why this? Which are the mechanisms? How indicate? Also, a long sentence that is hard to follow.

 

l.381: Please provide the links to the data sources.

 

References

l.395: Science with a capital letter.

l.457: Some references have DOI while others do not. Please check the journal guidelines.

l.466-467: Please check the capital letters.

l.482: Science with a capital letter.

 

l.482: …-2000-13”?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 5 Report

The study by Hu and Lane is presenting the results of both clean and debris-covered glaciers on the Tibetan Plateau based on Landsat imagery. I highly welcome the effort of the author to improve on the currently available datasets for the region. Overall, it is very well written manuscript. Below I provide my minor comments that could improve the manuscript before final publication.

 

 Please use retreat rate of −0.0 % yr⁻¹ instead of 0.0 %/a everywhere in the manuscript

 

You could cite Zemp et al., 2019 and Hugonnet et al., 2021 in the first paragraph of the manuscript

 

Page 2 Line 51 ….[9,10]…. – you could cite Xie et al., 2022 even it is under discussion now

 

Page 2 Line 66 … [18] … - please cite Bolch et al., 2010 here

 

Page 2 Line 70 …. [30,31] … - please also cite Holobâcă et al., 2021 here

 

Page 3 Figure 1 – regional insert map should be added

 

Page 3 Line 94 - …Landsat series of images from 1985-2020 were used

Please specify which Landsat images. 5? 7? 8?

 

Page 3 Line 97-98 - …Digital elevation was obtained from The Shuttle Radar Topography Mission (SRTM), …

From which year? please specify

 

Page 10 Figure 10. – can you provide panel B for summer temperatures and winter precipitations along with an appropriate analysis in the text?

 

Page 12 Chapter 5.2.2 Change rate of glaciers

I think you need more discussion and explanation here why “The clean glaciers showed an obvious retreat trend, with an average annual retreat rate of 0.55%/a, while the area of debris-covered glaciers showed an expanding trend, with a change rate of 0.62 %/a, but with large variability in the rate of change over time.” that you state in the Abstract and Conclusion.

In the Discussion section please take a chance and compare your results with results from other regions, such as Caucasus Mountains (Tielidze et al., 2020), Karakoram and Pamir (Mölg et al., 2018), etc. This will allow you to create a big picture rather than just focus on a regional scale. It also makes your research more readable and cited by the international community.

 

I hope the authors agree with these recommendations and can implement the suggested improvements.

 

Reference

Bolch, T., Menounos, B., and Wheate, R.: Landsat-based inventory of glaciers in western Canada, 1985–2005, Remote Sens. Environ., 114, 127–137, https://doi.org/10.1016/j.rse.2009.08.015, 2010. 

 

Holobâcă, I. H., Tielidze, L. G., Ivan, K., Elizbarashvili, M., Alexe, M., Germain, D., Petrescu, S. H., Pop. O. T., and Gaprindashvili, G.: Multi-sensor remote sensing to map glacier debris cover in the Greater Caucasus, Georgia, J. Glaciol., 67, 685–696, https://doi.org/10.1017/jog.2021.47, 2021. 

 

Hugonnet, R., McNabb, R., Berthier, E., Menounos, B., Nuth, C., Girod, L., Farinotti, D., Huss, M., Dussaillant, I., Brun, F., and Kääb, A.: Accelerated global glacier mass loss in the early twenty-first century, Nature, 592, 726–731, https://doi.org/10.1038/s41586-021-03436-z, 2021.

 

Mölg, N., Bolch, T., Rastner, P., Strozzi, T., and Paul, F.: A consistent glacier inventory for Karakoram and Pamir derived from Landsat data: distribution of debris cover and mapping challenges, Earth Syst. Sci. Data, 10, 1807–1827, https://doi.org/10.5194/essd-10-1807-2018, 2018.

 

Tielidze, L. G., Bolch, T., Wheate, R. D., Kutuzov, S. S., Lavrentiev, I. I., and Zemp, M.: Supra-glacial debris cover changes in the Greater Caucasus from 1986 to 2014, The Cryosphere, 14, 585–598, https://doi.org/10.5194/tc-14-585-2020, 2020.

 

Xie, F., Liu, S., Gao, Y., Zhu, Y., Bolch, T., Kääb, A., Duan, S., Miao, W., Kang, J., Zhang, Y., Pan, X., Qin, C., Wu, K., Qi, M., Zhang, X., Yi, Y., Han, F., Yao, X., Liu, Q., Wang, X., Jiang, Z., Shangguan, D., Zhang, Y., Grünwald, R., Adnan, M., Karki, J., and Saifullah, M.: Interdecadal glacier inventories in the Karakoram since the 1990s, Earth Syst. Sci. Data Discuss. [preprint], https://doi.org/10.5194/essd-2022-265, in review, 2022.

 

Zemp, M., Huss, M., Thibert, E., Eckert, N., McNabb, R., Huber, J., Barandun, M., Machguth, H., Nussbaumer, S. U., Gärtner-Roer, I., Thomson, L., Paul, F., Maussion, F., Kutuzov, S., and Cogley, J. G.: Global glacier mass changes and their contributions to sea-level rise from 1961 to 2016, Nature, 568, 382–386, 2019. 

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Hu et al.’s resubmission has addressed many of my concerns from the earlier version. I have a few remaining questions and one large concern. Additionally, I have provided a few line-specific comments.

 

 

General Comment/Concern: 

 

I appreciate that you have remade the glacier area change figure as time series and added regression analyses to them. However, how you analyzed the data has a large issue that needs to be addressed. That issue is how you handle the time bins (i.e., your x-axis). I am not sure of the best method to address this issue, but I believe how you addressed it is incorrect. I would recommend asking a colleague in the statistics department for guidance.

 

I believe there is an issue with how you treat the time intervals. Since you don’t have a single measurement per year but instead one measurement for multiple years, the regression analyses may be trickier. Based on the figures, it looks like you created a data point for each year in an interval. For example, your first measurement is from the 1985 to 1993 time window. In your plots, you created 9 data points with the same y-value (the composite area for that interval). In your regression analysis, however, that one measurement from the 1985 to 1993 interval now plays an oversized role. In reality, you should have a single measurement from 1985 to 1993 with large temporal uncertainty.

 

I speculate you’ll need to add weightings to your regression analysis. The later intervals, which are more temporally constrained, should be weighted heavier than the earlier intervals, which are less temporally constrained. At present, however, your method does the opposite. The earlier intervals, which are less temporally constrained, are affecting the regression analysis more than your later intervals, which have better temporal constraints. A colleague in the stats department woudl likely know an easy way to address this issue.

 

 

 

Specific Comments:

 

Line 24: clean glaciers retreat fast  clean glaciers retreat faster (or maybe its fastest?)

 

 

Line 66: The phrase that starts 'Because the spectral information …' is not a complete sentence.

 

 

Line 191 & 196: How do your average retreat rates found by differencing the areas in the earliest and latest intervals compare to the slopes of your linear regression lines in Figure 6? A value of the linear regressions from your area loss time series plots is that they provide an alternative way to figure out the retreat/growth rates.

 

 

Line 228: Does the phrase ‘similar to normal’ indicate that you ran tests and determined that one (or both) is (are) distinguishable from a normal distribution?

 

If you haven’t run tests for normality, I would still encourage you to do so. There are a variety of ways to estimate the tendency with normality, from visual plots (beyond just the frequency distribution curves) to calculating the third (skew) and fourth (kurtosis) moments. There are more rigorous tests too, but those are likely not needed.

 

 

Line 248: There may be a typo or missing word. When I look at Figure 9b, it doesn’t seem to me that the greatest growth rate for debris cover glaciers happens at 20 deg.

 

 

Figure 10b: Replot the rose diagram so that the radius axis has only positive values. If you make the radial axis the retreat rate (i.e., -1 x change rate), then the aspect with the highest retreat rate will be where the curve is furthest from the origin. Looking at this plot quickly, I concluded that the southwest aspects are the ones that are changing the most since they are the ones where the curve is furthest from the origin. In my first read of the text, I thought your text contradicted the data in the figure. It wasn’t until I went back and put some work into reading the figure that I understood what it was trying to present. As created, this plot will be very opaque to many readers. This issue can be addressed by making the radius values in the middle figure positive.

 

 

Table 3/Line 275-276: Either in the table or in the text, present the p-values. I’m left asking, especially for the winter precipitation, whether the trends are significant. You could add a p-value column to the table or make a note in the table label (and bold/asterisks) the rows where there are statistically significant p-values. Or you could note in the text that they are significant (if that is the case) and what p-value you used for significance.

 

 

Line 335 - 337: Be cautious saying that debris-covered glaciers are expanding. That makes me think that you are saying debris-covered glaciers are advancing downhill. If that is the case, that would be really interesting. However, your results from above can’t tell us if there has been an advance of debris-covered glaciers.

 

What you can say is that there has been area loss in clean-ice glaciers while debris-covered glaciers have gained area. You hint in the succeeding lines about why we might observe these changes: some of the clean ice loss could be due to debris cover on formally clean ice.

 

You have the data to explore this question! A nice addition to your analyses/discussion could be seeing what clean-ice glacier area loss was due to pixels being classified more recently as debris-covered glaciers. Similarly, you can see which debris-cover glacier area gain is due to the re-classification of clean-ice glacier area versus which (if any) is due to the advance/expansion of a debris-covered glacier from the 1985 to 1993 interval.

Author Response

Please see the attachment.

Author Response File: Author Response.docx