Glacier Change in the West Kunlun Main Peak Area from 2000 to 2020
Round 1
Reviewer 1 Report
Major concerns
1) The authors used the Bahr's power-law relationship (i.e. V-A scaling) to derive the ice volume from glacier area. It is generally OK to calculate the regional glacier volume using the V-A scaling relationship. But according to Bahr et al. (2015), the V-A scaling relationship should be avoid applying to glacier complexes. It means that the V-A scaling relationship is not suitable for glaciers with multiple branches, while the glaciers in West Kunlun Shan mostly have multiple branches.
2) To calculate the volume change, geodetic volume change derived from DEM differencing is more precise and reliable than the V-A scaling relationship. In some cases, the glacier area and terminus generally remains stable, but the downwasting causes the glacier mass loss. So it is not appropriate to derive the glacier volume change from area change in West Kunlun.
3) In Section 3.4, the authors tried to get statistical relationships between glacier length and area (perimeter). What are the significances of obtaining such relationships? Can we use these relationships in other glacierized areas? If these relationships were just local or suitable for the west Kunlun, do we need such relationships? The glacier length is calculated from the main trunk of the glacier, but the area calculation includes the areas of tributaries. My another concern about the statistical relationships is the justification of such relationships derived from multi-branches glaciers.
4) This manuscript suggests that the lag time of glacier response in the West Kunlun is 14 years. From the results of glacier area change, we can see that the responses of glacier area changes are heterogeneous. So if the 14-year lag time is a mean value?
5) Does the lag time equal to the glacier response time? The glacier area maybe not a good indicator for glacier response to climate change. As I stated before, area and terminus of some glaciers may remain stable, but such glacier are losing/increasing mass. So the lag time may not a real indicator for glacier response.
6) Through the study of glacier area change in the West Kunlun over the 21st century, can we get some insights about the Karakoram-West Kunlun glacier anomalies? Do the anomalies disappear?
General comments
The subsection 2.1 Study Area should be defined as a new section independent on the current Section 2.
The Figure texts should be improved and be more clear, particularly in Figure 15.
Specific comments
line 56: The abbreviation of WKL is first occurred, so its full name should be given here.
lines 58-59: Which several studies? Citations should be inserted here.
line 59: Researches on glacier area has…… ---> Studies on glacier area have……
lines 63-64: "Research on …,ICESat data." Citation should be inserted following this sentence.
line 68: What are h methods?
lines 75-80: You give long descriptions about the background and objective of the 3rd Xinjiang scientific investigation project. It's not necessary to introduce it here. Because it is just a comprehensive project but not the scientific problem itself.
line 111: a fifth-generation ---> the fifth-generation
line 133: what is Cy4lr2?
line 144: Red --> red
lines 172-173: The C and γ values should be given here.
lines 180-182: In Equation 2, you indicate that N is the perimeter of the glacier outline, and A is the half pixel. However, in Equation 6, you use the symbols P and Hc to represent the perimeter and half pixel. The symbols should be kept consistent in this manuscript.
line 201: It should be justified that the methods about lake and glacier area changes are same.
lines 238-239: change " we observed an increase in glacier areas with an area between <0.1 km2 and 50-100 km2" to " we observed glacier area increasing for area class <0.1 km2 and 50-100 km2”
line 445: Glacier division ---> Glacier separation
line 492: I think this study just investigated the geometrical change characteristics of West Kunlun glaciers, not the dynamic change characteristics (flow speed change, geodetic elevation change, stress regime change, etc.)
line 511: what is glacier treat?
References:
Bahr, D. B., Pfeffer, W. T., & Kaser, G. (2015). A review of volume-area scaling of glaciers. Reviews of Geophysics, 53, 95–140.
This manuscript is generally well-structured. But the language should be improved by native speakers.
Author Response
The reply to the review comments is at the back of PDF. Please pay attention to check. Thank you!
Author Response File: Author Response.pdf
Reviewer 2 Report
See pdf
Comments for author File: Comments.pdf
Author Response
The reply to the review comments is at the back of PDF. Please pay attention to check. Thank you!
Author Response File: Author Response.pdf
Reviewer 3 Report
The authors use Landsat data from 2000 to 2020 to track glacier changes in the West Kunlun Main Peak Area of the Tibetan Plateau region. Glaciers in this sub-region of the Tibetan Plateau region have not changed much, which is different from the broader trends in the region. This glacier behavior has been deemed the Karakoram anomaly. The study has compiled a rich dataset of glacier areas in the region from 2000 through 2020. It is at a higher temporal resolution than previous studies. With remote sensing measurements of glacier areas, the authors calculate the lengths and volumes of the glaciers. These glacier variables are then analyzed using a range of descriptive statistics and some empirical models between variables. The data collected/calculated are meaningful, and the analyses seem well conducted. However, the study has a fundamental issue: the results need to be adequately considered in relation to the error/uncertainty in the data.
Many of the results analyzed are not statistically distinguishable from no change. For example, in Table 1, there have not been statistically resolvable glacier changes in this region. The authors need to state as much about their results. The lack of resolvable changes in region-wide glacier area (and likely length and volume, although the uncertainty/error for these variables is not included in the tables) renders moot much of the data presented in Section 3.1. Similarly, in Sections 3.2 and 3.3, many of the changes are very small and likely within the realm of uncertainty/error in the data. However, the uncertainty/error values are not included in the text/visuals, precluding such an analysis by the reader. Without assessing the signal of changes versus uncertainty/error in the data (i.e., the noise), it is impossible to determine which discussion points are valid or not. However, given the small magnitude of changes change and typical uncertainty values, I suspect that many of the changes discussed and used are not statistically resolvable from zero.
Despite these critiques, there are many good things about the data and analyses of this study. At a minimum, the authors need to clarify which findings are significant. The authors should ensure that only significant results are included in the discussion sections and the summary sentence(s) of the results sections. In addition, the uncertainty/error values should be calculated for all the variables and included in the presentation (both in text and visuals). Below I provide some general comments for further consideration in a revision. I also provide some line-specific suggestions to improve a revised manuscript.
General Comments For Revision:
- The lack of statistically resolvable changes in this region seems consistent with previous studies of the Karakoram anomaly. Thus, part of this study could be a confirmation of this previous finding, using newer and more data.
- Also, the focus in the paper could be shifted more toward the few glaciers with distinguishable changes. Figure 7 illustrates that some glaciers in the region are changing. Identifying which glaciers are changing and understanding why they are changing (while the others are not) could be a fruitful direction of focus.
- When interpreting results, error/uncertainty intervals should be included: both in the text and the figures. For example, adding uncertainty intervals in Figure 9 would help identify glaciers with meaningful changes. Also, the trend in Figure 9 may be due mainly to the few glaciers with meaningful changes, while the rest may have an R-value close to zero.
- I feel that the method for calculating the uncertainty/error of glacier area may underestimate its value. I provide some references below for alternative methods to calculate uncertainty/error. Also, error/uncertainty values should be calculated and added in for the length and volume.
- Figures with multiple subplots of the same variables but different intervals of time (e.g., Fig. 4, Fig. 5, Fig. 6, Fig. 13, Fig. 16) should have the same y-axis range as the figures for the other intervals. By changing the intervals, it is harder to make comparisons between plots. Also, it makes it easier to over-interpret small changes in one interval if it has a small y-axis range but under-interpret small changes in another that has a wide y-axis range.
Introduction, Suggestions For Clarity & Typos:
- Line 35: for brevity, ‘The cryosphere is the sphere that undergoes the most significant changes’ ‘The cryosphere is undergoing significant changes’.
- Line 35: While I agree that it is undergoing significant changes, I am not sure if it’s the zone with the most significant changes.
- Line 47: I found the first paragraph hard to follow. There are multiple key ideas in it. These ideas could be presented more clearly in shorter paragraphs, each focusing on a key idea (or two). Line 47 could be a good location for a paragraph break. At Line 47, the focus shifts from climate and glacier changes more generally to their specific changes in the Tibetan Plateau and Pamir-Karakoram-West Kunlun region.
- Line 56: The acronym WKL is not defined. I also don’t see if used elsewhere in the paper. I recommend spelling out the name in full. That way the reader would only need to recall one acronym: WKMPA.
- Line 56: for brevity, ‘According to the China Second Glacier Inventory (CSGI) [19], The WKL [spell-out name] region is the most concentrated area of large mountain glaciers in China, attracting … . ‘ ‘The WKL [spell-out region] is the most concentrated area of large mountain glaciers in China [19], attracting … .’
- Line 59: ‘Researches’ ‘Research’
- Line 61: The acronym WKMPA is not defined in the main text (only the abstract). My understanding is that it is best practice to redefine it in the Introduction if it is defined in the abstract. As a reader, I appreciate having an unfamiliar acronym defined a few times in the text!
- Line 71: I found it hard to keep track of all the key ideas in this paragraph. The concluding note in Line 71 could act as a nice transition to a new paragraph. That new paragraph could open as follows: ‘Previous studies of glacier change in the MKMPA region have mainly focused on individual factors such as … . However, given the current state of glacier anomalies in the region, it is of great significance to …. ‘
- Figure 1: An regional map insert, framing this location within greater High-Mountain Asia would help readers orient themselves.
Methods Details That Should Be Added In Text/Table(s) For Reproducibility:
- Line 124: What pan-sharpening algorithm was used? And how was it conducted? If it was one of the standard ones in ArcGIS, then I would state as much. I’d also recommend at least mentioning the name of the algorithm that ArcGIS (or another GIS package) is using.
- Line 130: I believe ERA5-Land is being used in this analysis, given the spatial resolution of 0.1 deg. The resolution of ERA5 (atmosphere) is 0.25 deg.
Line 146: I feel more discussion is needed about the band ratio threshold values. Was a single value used or all scenes? Or was a unique threshold found for each scene? How was (were) the threshold value(s) determined? Also, the threshold value(s) should be included in the text (or a table, which could be added to the supplemental/appendix) so that the results of this study are reproducible.
Line 155: ‘glacier terminus’ ‘glacier’s terminus’
Line 156: ‘glacier inventory’ ‘glacier inventories’
Line 170 (Equation 1): I feel more discussion is needed about the empirical parameter values in Equation 1. Are they the same for all glaciers (and all years)? And if so, then these choices should be discussed. Also, the value should be included somewhere in the document. If a constant value is used for glaciers, then the two parameter values could be included in the body of the text. If multiple values are used, then a table (either in the appendix or supplementary) could present these values effectively.
Line 170 – 169: I feel more discussion is needed on how the volume was calculated. Was the volume calculated for each individual glacier, using the empirical relationship in Equation 1, and then summed together to find the total volume? Or was the area of each glacier first summed to find a total area and then that area was used in Equation 1 to find the volume? My understanding is that the empirical relationship in Equation 1 is for a single glacier. Therefore, the first approach would be more correct: finding the volume for each glacier and then summing the volumes for the glaciers to find a total volume. Unless the gamma parameter is 1, the two above-mentioned methods would yield different total volumes.
Line 203 (Equation 6): Equation 6 and Equation 2 seem related to me. However, as written, there are different symbols representing the same quantity. This obscures the relationship between these equations. I believe that deltaE could be rewritten as: deltaE = 1/2 (Ei + Ej), where Ei is the error for year i (using Equation 2) and Ej is the error for year j (using Equation 2).
Methods: Error/Uncertainty In Measurements
Line 175: Regarding errors in classifying glacier extents, I recommend Hanshaw & Bookhagen (2014) and Kochtitzky et al. (2018). Both provide clear discussions of sources of uncertainty when working with data from recent Landsat missions. They also provide nice analytical expressions to quantify uncertainty/error using information obtained in polygons/shapefiles. I have provided the bibliographical information for each reference at the end of the document.
Line 182: Why is A equal to ½ the pixel size? Has this method been used in previous studies? If so, please cite that work to justify for this approach. However, it is a different approach than I have used for such applications. I have seen different approaches used to address the question of area uncertainty from remote sensing data. One approach, similar in spirit to the one used in this study but quantitatively different, assumes that the maximum uncertainty occurs when all pixels are misclassified. For this approach, the uncertainty is E = n x G, where G is the length of the grid cell (i.e., 30 meters for Landsat 4/5 scenes and 15 meters for pan-sharpened Landsat 7/8/9 scenes). A nice discussion of this approach is found in Kochtitzky et al. (2018). Another approach also assumes that uncertainty stems from the misclassification of pixels along the parameter. But the second approach assumes that misclassified pixels are normally distributed. A nice discussion of the second approach is found in Hanshaw & Bookhagen (2014).
Line 211: Is there a method to quantify the error/uncertainty in the glacier volume? Since the volume depends on the area (according to Equation 1), error/uncertainty in the area would propagate to error/uncertainty in the volume. If there is not an analytical expression (like in Equation 2 or 3), then the error/uncertainty in glacier volume could be bootstrapped. Equation 1 could be used to calculate the volume for each glacier multiple times using different values of its area (within the range of possible values). This step could be repeated 100 (or 1,000) times for each glacier, and the mean and variability in volumes could be quantified. This bootstrapped approach could provide a quick estimate of the uncertainty in the glacier volume.
Results Comments:
Line 214: ‘Table 1 presents the number, area, length, and ice volume of glaciers in WKMPA ..’ ‘Table 1 presents the total number, area, length, and ice volume of glaciers in the WKMPA ..’
Line 213 - Line 229: This analysis does not account for uncertainty/errors. The relative changes and temporal variations in their values do not seem to exceed the uncertainty/error in the data. Using the data in Table 1 and the criteria for change in Equations 4 through 6, I find that net change is much smaller than the uncertainty/error: Abs(delta S) << delta E. I cannot perform similar analyses for length or volume since uncertainty/error values are not provided in Table 1 (see comment about Line 230 (Table 1 values) below). This analysis should be repeated for the glacier length and volume to see if changes are significant. This finding, however, is not bad. The data in this study support the findings of previous studies in the region pertaining to the Karakoram anomaly [14-18].
Line 230 (Table 1 caption): Make explicit that these values are the totals for the region.
Line 230 (Table 1 values): Uncertainty/error ranges should be added to length and volume.
Line 238: ‘area class’ –> ‘area classes’
Line 251 (Figure 4a, composition): The y-axis for the number bar charts (and area line graphs) crosses an x-axis value below 0. That makes this figure very misleading and hard to interpret.
Line 251 (Figure 4b-g): The different ranges of y-axis values make comparing the plots challenging.
Line 272 (Figure 5b-g): The different ranges of x-axis values make comparing the plots challenging. See the comment above for Line 252 (Figure 4b-g).
Line 297 - Line 297: I believe there is a typo in this line. The topic sentence describes another type of Figure than Figure 7.
Line 300: km^2 not superscripted
Figure 7: That’s a great figure!
Line 333 – Line 345 / Figure 9: Interesting, there aren’t really any points in the 2nd quadrant. Any that are so close to the origin that they are likely indistinguishable from zero.
Line 348 - 349 / Figure 10: Re-bin the figure to more align with Figure 7. There is some interval of length change that is indistinguishable from no change. That bin could be green, as in Figure 7. At present, the bins 0 to 0.5 bin and -0.5 to 0.0 bin include both instances of no change and some growth and decay, respectively. As binned, however, it is hard to figure out which glaciers aren’t changing notably.
Discussion Section:
Line 377: ‘Hydrothermal’ ‘air temperature and precipitation’; hydrothermal makes me think of underwater vents associated with sea-floor spreading.
Line 381 - Line 383: Make explicit that this statement about climate forcings and time scales is true for the study region; that statement is not true for all glaciers.
Table 2: How were the climate trends calculated? Were the data points for the 5 years plotted and a trendline fit to them? If so, discuss this process. If, instead, the trends come from just the difference in the first and last year, then that would be a fraught method to figure out the 5-year trend.
References With Good Approaches To Glacier Uncertainty/Error:
Kochtitzky, W.H., Edwards, B.R., Enderlin, E.M., Marino, J. and Marinque, N., 2018. Improved estimates of glacier change rates at Nevado Coropuna Ice Cap, Peru. Journal of Glaciology, 64(244), pp.175-184.
Hanshaw, M.N. and Bookhagen, B., 2014. Glacial areas, lake areas, and snow lines from 1975 to 2012: status of the Cordillera Vilcanota, including the Quelccaya Ice Cap, northern central Andes, Peru. The Cryosphere, 8(2), pp.359-376.
Overall, the writing was clear. There were only a few typos or areas for increased clarity, which were noted in the comments above.
Author Response
The reply to the review comments is at the back of PDF. Please pay attention to check. Thank you!
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
lines 8-9: studying on their dynamics…… ---> investigation of their dynamics……
line 56: delete ", attracting the attention of many scholars"
line 72: factors ---> characteristics
lines 72-73: long data intervals ---> long time intervals
lines 75-77: delete the sentence "To better……in July 2019 "
line 90: Basin ---> basins
lines 506-507: It's difficult to understand this sentence. Which parameters can we calculate from the relationship between different glacier parameters?
line 509: delete "two-dimensional perspective"
line 511: how can you get that the glacier area will increase in the future? Will the future temperature decrease?
The English writing should be improved by native English speakers. As I'm not a native English speaker, I could not help more to improve the language.
Author Response
1) lines 8-9: studying on their dynamics…… ---> investigation of their dynamics……
Response:
Thanks to the reviewers for their comments.
The section has been revised and we have adjusted it to “Glaciers are sensitive indicators of climate change, and investigation of their dynamics is crucial for ensuring regional ecological security as well as disaster prevention and mitigation measures.”.
2) line 56: delete ", attracting the attention of many scholars"
Response:
Thanks to the reviewers for their comments.
The section has been revised and we have adjusted it to “The West Kunlun region is the most concentrated area of large mountain glaciers in China [19].”.
3) line 72: factors ---> characteristics
Response:
Thanks to the reviewers for their comments.
The section has been revised and we have adjusted it to “Previous studies on glacier change in this region have mainly focused on individual characteristics such as area, surface elevation and movement velocity with relatively long time intervals.”.
4) lines 72-73: long data intervals ---> long time intervals
Response:
Thanks to the reviewers for their comments.
The section has been revised and we have adjusted it to “Previous studies on glacier change in this region have mainly focused on individual characteristics such as area, surface elevation and movement velocity with relatively long time intervals.”.
5) lines 75-77: delete the sentence "To better……in July 2019 "
Response:
Thanks to the reviewers for their comments.
We deleted it in the revision.
6) line 90: Basin ---> basins
Response:
Thanks to the reviewers for their comments.
The section has been revised and we have adjusted it to “The Kunlun Mountains traverse the northern region of the Tibetan Plateau, stretching from the Pamir Plateau in the west to the Ruoergai Basins in the east. It borders on the hinterland of the plateau, and is adjacent to Tarim and Qaidam Basins in its northward direction [26].”.
7) lines 506-507: It's difficult to understand this sentence. Which parameters can we calculate from the relationship between different glacier parameters?
Response:
Thanks to the reviewers for their comments.
The section has been revised and we have adjusted it to “(3) It’s worthwhile to explore the relationship between different glacier parameters to help us calculate the other parameters quickly to some extent.”.
What we mean here is that the length of a glacier can be quickly estimated from its area and perimeter. In addition, we would try to introduce more glacier parameters to correct the existing empirical formulas in the future study.
8) line 509: delete "two-dimensional perspective"
Response:
Thanks to the reviewers for their comments.
The section has been revised and we have adjusted it to “(5) The mechanism by which glacier division affects glacier change is more complex. In general, it is necessary to analyze its changes concretely before and after division.”.
9) line 511: how can you get that the glacier area will increase in the future? Will the future temperature decrease?
Response:
Thanks to the reviewers for their comments.
As we mentioned in this paper, rising temperature is the main cause of the decrease in glacier area. Correspondingly, decreasing temperature would lead to a positive mass balance of the glacier, which will result in an increase of glacier area. In Section 5.1, the response of glacier area to climate change, we get that a 14-year time lag between temperature changes and glacier area response in this region. Therefore, we can predict the trend of glacier change in this region during 2020-2034 based on the temperature change during 2006-2020. It is therefore questionable whether the wavelet analysis in Figure 13 is actually suitable for a forecast. Accordingly, there should be another period of cooling. Based on the temperature change trends in Figure 12 and Figure 13, we can see that the temperature change in this region during 2006-2020 would firstly decrease and then increase, and the corresponding glacier area change should firstly increase and then decrease. Especially in a short period of time after 2006, the temperature showed a downward trend in a short period. Therefore, we can conclude that the glacier area would increase in the future for a short period of time, and then decrease.
However, it is worth nothing that the conclusion that the glacier area in the study area would increase in a short period was based solely on the temperature and precipitation trends in the region, and didn’t take into account global climate change. Therefore, this conclusion is currently controversial, and we deleted it in the revised draft.
Author Response File: Author Response.docx
Reviewer 2 Report
See pdf
Comments for author File: Comments.pdf
Author Response
Report remotesensing-2471680-peer-review-v2
Glacier change in the West Kunlun Main Peak Area from 2000 to 2020
1) After the revision, the first version of the paper by authors Cong Zhang, Xiaojun Yao, Suju Li, Longfei Liu, Te Sha, and Yuan Zhang was significantly improved. This applies to both the formal presentation and the content. The reviewer's comments were essentially taken into account and/or explained in the cover letter, even if this was not entirely satisfactory in individual cases. But maybe this is not so easy to do. A discussion of the impact of the determined error range on the analysis would have been desirable.
In the revised version, the authors refrain from considering changes in volume of the glaciers and limit themselves to the results of changes in area and length. Since, depending on the size of the glaciers, their ice dynamics modifies the reaction of the glacier geometry to the climate signal to a considerable extent, this does not provide a clear relationship between glacier behavior and climate trends. However, the authors confine themselves to presenting the changes in geometry determined with their remote sensing methods, which, however, are only of minor statistical significance in view of the error margins and the relatively short investigation period of 20 years.
Nevertheless, the analysis carried out appears to be useful, especially since the West Kunlun Main Peak Area is apparently located in a region that is one of the rare regions with a significantly lower degree of climate change. This is not well known, given that China generally has an above-average rate of temperature change. The newly designed Figure 1 now illustrates the larger-scale location of the study area in the south-west of the country, which in recent decades has actually shown lower warming trends at 0.15°C per decade than in north-east China at 0.4°C per decade.
This was not clear in the first version of the paper, so the redesigned Section 5.1 together with Figure 12 provides important information for understanding the results. Note that Figure 12a is a seasonal temperature curve. In large parts of the world, however, since 1980, after the minima in the phase of global dimming in the 1960s and 1970s, the strongest rise in temperature moreover the level around 1950 has been recorded, especially in summer. This base structure is also evident in the Kumlun Mountains, but the increase after 1980 is much smaller. In addition, the general period of less warming between 2000 and 2010 (“hiatus”) is much more pronounced. Since 2010, however, the temperature has been rising again. It would be helpful at this point to refer again to the literature reference [17], which provides further information on the temperature curve. This means that the results in Table 12 can also be understood.
There is much to debate about the cause of the anomaly, including whether it will persist over time. There has definitely been a warming trend in the Kumlun region over the last decade. Presumably this to continue given the global trend. It is therefore questionable whether the wavelet analysis in Figure 13 is actually suitable for a forecast. Accordingly, there should be another period of cooling.
However, this does not necessarily have to be as strong as that before. But this is only one opinion among many.
All in all, the article as it stands is ready for publication. The bibliography has also been revised accordingly.
Response:
I would like to thank the reviewers for taking the time to read this manuscript and put forward many corresponding comments. In the process of data processing, we strictly followed the process of China's second glacier inventory data, and performed fusion processing on Landsat OLI images, which doubled the spatial resolution compared with the previous one. So there is a more precise control of the error.
In addition, it is worth noting that the temperature rise rate in China you mentioned is relatively high. In contrast, the warming rate in the study area is not obvious. The response time of the glacier area in this region to the temperature change is 14 years, which corresponds to the forecast of the glacier area change in the future, and the temperature change after 2006 should be mainly considered. It is more obvious that the overall warming rate during 2000-2010 was small, and even the temperature decreased, while the warming rate during 2010-2020 was significant. This can be observed more clearly in the wavelet analysis in Figure 13. However, to predict the trend of glacier area change based on temperature change, our conclusion is that the glacier area change in this region would show a small increase and then a significant decrease trend in 2020.
2) Only the few passages listed below are still recommended for revision:
Line 166: „observed“, better „reviewed, checked“
Response:
Thanks to the reviewers for their comments.
The section has been revised and we have adjusted it to “We reviewed the four main sources of error in the process of the glacier mapping.”
3) Line 178: "artificial understanding difference" is incomprehensible. Is there a better clearer term?
Response:
Thanks to the reviewers for their comments.
The section has been revised and we have adjusted it to “manual visual interpretation.” This part mainly refers to the deviation of visual interpretation results due to the difference personal understanding.
4) Figure 4: The legend or the descriptive text is missing a note or an explanation of the error bars.
Response:
Thanks to the reviewers for their comments.
We added the legend and the descriptive text in the Figure 4.
5) The conclusion has been thoroughly revised. However, the statement in line 511/512 is questionable given the expected global warming trend.
Response:
Thanks to the reviewers for their comments.
As we mentioned in this paper, rising temperature is the main cause of the decrease in glacier area. Correspondingly, decreasing temperature would lead to a positive mass balance of the glacier, which will result in an increase of glacier area. In Section 5.1, the response of glacier area to climate change, we get that a 14-year time lag between temperature changes and glacier area response in this region. Therefore, we can predict the trend of glacier change in this region during 2020-2034 based on the temperature change during 2006-2020. It is therefore questionable whether the wavelet analysis in Figure 13 is actually suitable for a forecast. Accordingly, there should be another period of cooling. Based on the temperature change trends in Figure 12 and Figure 13, we can see that the temperature change in this region during 2006-2020 would firstly decrease and then increase, and the corresponding glacier area change should firstly increase and then decrease. Especially in a short period of time after 2006, the temperature showed a downward trend in a short period. Therefore, we can conclude that the glacier area would increase in the future for a short period of time, and then decrease.
However, it is worth nothing that the conclusion that the glacier area would increase in a short period was based solely on the temperature and precipitation trends in the study region, and didn’t take into account global climate change. Therefore, this conclusion is currently questionable, and we finally deleted it in the revised draft.
Author Response File: Author Response.docx
Reviewer 3 Report
I appreciate the time and effort the authors took to address my first round of comments. Many parts of the article have improved from the first round. One such area is the figures, which are clearer and less likely to lead a reader astray. Also, the authors have added new aspects of uncertainty into their analysis.
The revised manuscript, however, still does not adequately (and in some cases accurately) address the uncertainty/error of the data. While uncertainty/error values have been added to many visuals, the text of the findings only includes the central values for amounts/rates of change. The text does not include the uncertainties associated with the changes. In the instances where I could glean those uncertainties, they exceeded the magnitude of the changes presented in the text. Thus, multiple instances exist where the data do not support what is said in the text.
A fundamental challenge I (still) see is that region-wide changes are not statistically resolvable. That is also true for most individual glaciers. However, the results of the manuscript focus on describing in detail the small changes, which in many cases are not statically resolvable. These small changes, which are mostly not significant, are then used in the discussion to talk about important processes. However, in most cases, I don't see evidence that the null hypothesis (no change) can be rejected.
The revised manuscript has not changed meaningfully in its scope from the original submission. Since most of the measured changes are not significant, the scope is not appropriate. I believe a paper about non-discernable change can make a meaningful contribution. In this study, the author's findings support previous research about the area, using newer methods and more data. That finding can be important. Moreover, the authors have identified some glaciers changing within a mostly static region. That is also interesting. However, the manuscript continues to focus on telling a story about small changes, and these changes are often not statistically significant.
It seems that there is an interesting story that can be told about the data collected for this study. That story, however, needs to address uncertainty adequately. I encourage the authors to think about what information is presented, considering the small changes for the region and data constraints.
General Comment, Introduction:
The introduction is much clearer. I appreciate the changes made. It makes it easier to follow. Also, it is more engaging!
General Common On Diction:
In many places, the word significant/significantly is used. I believe it's being used to indicate these values are larger than the other discussion/region-wide average. However, significance is usually used when discussing results that surpass a threshold of a statistical test. For example, A and B are statistically different (i.e., the p-value is smaller than some pre-determined threshold for significance (e.g., 0.05) using a statistical test about two populations, such as a t-test). If statistical tests were not used to compare values, then I would advise again saying a value is significantly different.
General Comment, On Uncertainty:
The overall uncertainty <75 km2 is surprisingly small. The uncertainty in 2000 is ~2.4%, which is quite small in light of values found in many other studies: usually 5% to 10% (or more). Equation 2 has the mathematical form of how 1-sigma uncertainties are propagated. Are the uncertainties presented here the 1-sigma uncertainties? In that case, the 2-sigma uncertainty values are usually presented.
General Comment, Section 4.1 (General Change):
The 23.83 km2 of glacier area loss from 2000 to 2020 needs to be viewed in light of uncertainties. Using Equation 2, I get an uncertainty for the change of 82.20 km2, which greatly exceeds the magnitude of the change. Repeating that for the other time intervals, all the area changes are smaller than the uncertainty interval. Thus, it is not appropriate to state that the glaciers have changed over the entire interval (or any subintervals) or that there are changes in the loss rate over different time intervals. Any measured changes cannot be shown to differ from zero. Also, these changes are quite small (<25 km2) when compared to the scale of the glaciated area (>2,900 km2).
A similar analysis of the glacier length changes shows that nearly all of them are smaller than the uncertainty. If uncertainty in the change of length is propagated the same way as it is for the area (i.e., using Eq. 2), then all length changes are smaller than the uncertainty for change (84.8 m).
General Comment, Section 4.2 (Characteristics Of Glacier Are Change).
In the second half of paragraph 1, the newly added text (in red) contradicts what is shown in Figure 4. There are no discernable changes in Figure 4. In Figure 4b through 4g, the uncertainty intervals intercept with zero. In Figure 4a, the bar heights for the two intervals overlap in their uncertainty interval. Thus, what is said in the text is inaccurate from what is shown in the figure.
When discussing Figure 5, uncertainty should also be addressed. While there is no nice way to add an error bar to these types of plots, the values should be interpreted in light of typical uncertainty values. The largest bars for change I see are just shy of 10 km2. That interval has more than 400 km2 of glacier area. Thus, these changes represent at most 2.5%. That bar might pass the threshold for uncertainty. However, most of the other bars for change do not. The text discussing changes by elevation should consider the 'significance' of these small changes.
General Comment, Section 5.1
I appreciate the clarification on the method for including the climate change rate (and Figure 12e,f). However, I still have a question. My understanding is that a five-year sliding window was used to calculate a change rate. Are only change rates where the slope is statistically significant shown in Figure 12e,f)? If the p-value for the linear model is greater than the threshold (say 0.05), then the slope should be zero. Looking at Figure 12a,b, I am struggling to recreate the slopes shown in Figure 12e,f (by eyeballing local trendlines over 5-year intervals). Please clarify. And if the significance of the climate slopes was not considered, please reconsider the method.
A pressing concern is whether there are different glacier rates to try to explain climatically. In Table 2, uncertainty intervals must be added to the Change rate column. And if those uncertainty intervals mean the change rates are not distinguishable from one another (or from zero), then the analysis in Section 5.1 seems inappropriate. Without discernible changes (and changes discernable from zero), any connections identified between glacier changes and climate forcings are likely spurious.
Is there a reference to another paper that has used this approach for the time lag method? I am still not sure that this is an appropriate approach. Was a linear regression model with a lag used and a 14-year lag was found to be statistically significant and the lag with the best r2 value?
General Comment, y-axis Ranges In Figure 14 & Figure 17:
I appreciate that Figures 4, 5, and 6 were reformatted to have the same y-axis range. The response to my comments indicated that was done for these two figures, but they still have different y-axis ranges. In one case (Figure 14f), the y-axis range is <0.4 km2, which makes for impressive-looking glacier length fluctuations. However, this glacier is almost 200 km2 large, making the fluctuations probably not significant.
Specific Comments:
Line 61: The change from Reference 21 is expressed as a percent and not as a rate (% a-1). I am unsure if it's a typo or if the findings from Reference 21 are the total change given as a percent. If it's the latter, then convert the findings from Reference 21 into a rate by dividing it by the time interval of their study.
Figure 1: The data source for the glacier polygons should be included in the caption. I assume this is from an earlier (and maybe larger regional) study. If so, cite it in the figure.
Line 134-146: 'The range of threshold value ranged from 1.8 to 2.0' --> 'The threshold values ranged from 1.8 to 2.0'.
Figure 3: This is a great figure, but I didn't realize immediately that it is for two different methods. White space should be added between the glacier delineation and length extraction methods. Even better could be to break it up into (a) and (b) parts.
Line 173: 'large' --> 'larger'
Line 196: Please clarify what is meant by 'number of glacier length'. Is that the actual glacier length measure (e.g., 600 m)? If so, n seems like a strange variable to use. Or is n the number of pixels included in the length (e.g., 20 pixels for a 600 m long glacier)?
Line 214 & Table 1: It should be noted that it is the total glacier area and the average glacier length.
Line 252: Identify the statistical test(s)/analysis(es) used.
Figure 8a: Error bars are needed for the relative change rates to determine which (if any) of the values are distinguishable from zero.
Line 336: ‘terminal’ --> ‘terminus’
Line 343: I don't think this text has been updated considering the new bins in Figure 10.
Figure 10: I appreciate the changes, but some information has been lost. A three-bin range makes sense for Figure 7. If that same framework is used for Figure 10, are all length changes between -0.5 and 0.5 % a-1 indistinguishable from no change? I feel like the green bin is maybe too 'wide'. Perhaps a change of -0.4 % a-1 is meaningful. But if that is the case, such a change would be lumped into the green bin, which seems to represent no change. Maybe a 5-bin system would be better. There could be a light red and dark red bin for slightly significant and very significant retreats, respectively. The same could be used for advances but light blue and dark blue. I am not sure what would be the cutoff between no change and a slightly significant retreat/advance.
Line 454 - 460 + Figure 16:
From where did the pre-2000 glacier data come? Were these findings from another study? If so, that study should be cited. Or, is this additional analysis that used earlier Landsat scenes than were used elsewhere in the study? If it's the latter case, then the earlier Landsat data must be identified and presented in the Methods.
I have provided a few line-specific corrections in the Specific Comments section above. There is also a general comment on world choice (i.e., diction), pertaining to the use of significant/significantly.
Author Response
1) I appreciate the time and effort the authors took to address my first round of comments. Many parts of the article have improved from the first round. One such area is the figures, which are clearer and less likely to lead a reader astray. Also, the authors have added new aspects of uncertainty into their analysis.
The revised manuscript, however, still does not adequately (and in some cases accurately) address the uncertainty/error of the data. While uncertainty/error values have been added to many visuals, the text of the findings only includes the central values for amounts/rates of change. The text does not include the uncertainties associated with the changes. In the instances where I could glean those uncertainties, they exceeded the magnitude of the changes presented in the text. Thus, multiple instances exist where the data do not support what is said in the text.
Response:
Thanks to the reviewers for their comments.
We have re-supplemented the errors and uncertainties in the data in the text of the revised draft. Please refer to the revised draft for details. However, it is worthwhile that when we did relevant researches of glacier changes, we usually took the actual value we interpret into the calculation process, rather than the value after adding or subtracting the error. Therefore, we re-supplemented the error information in the returned document, but didn’t modify the value in a large scale. Please understand.
2) A fundamental challenge I (still) see is that region-wide changes are not statistically resolvable. That is also true for most individual glaciers. However, the results of the manuscript focus on describing in detail the small changes, which in many cases are not statically resolvable. These small changes, which are mostly not significant, are then used in the discussion to talk about important processes. However, in most cases, I don't see evidence that the null hypothesis (no change) can be rejected.
Response:
Thanks to the reviewers for their comments.
In the previous researches of glacier change, we usually interpreted the glaciation region based on remote sensing images to obtain the glacier contour data in different periods. The relevant parameters (area, length, ice volume, etc.) were analyzed with geographical significance in order to understand the glacier change in the study area and its response to climate change. Statistical methods were inevitably used in this process, but it is worth noting that we analyzed the factors that may have contributed to the interpretation error before we analyzed the specific changes. In addition, when comparing glacier changes in different regions, we usually took the real value for relevant calculation and supplement the error information, rather than bringing the error directly into the calculation, so as to make the data and relevant conclusions easy to understand.
The evaluation system and method of data error have been used in the related researches and verified. There is no doubt about the methodology. As for the statistical rejection of null hypothesis and other related principles you mentioned, the data in the manuscript is not enough to carry out statistical hypothesis testing and other related work. In addition, it is worthwhile that the study area in this paper can be regarded as a large area of glaciation, that is, ice sheet, and the distribution of glaciers is very concentrated. So, in the error assessment, as shown in the figure below, we only need to calculate the perimeter of the ice sheet, excluding the ridgelines. When calculating the glacier error in the previously more dispersed area, we need to consider the entire perimeter of the glacier. When we don’t consider the length of the ridgelines, the error of the glacier area should be 109.87 km2 (2000), 110.10 km2 (2005), 109.34 km2 (2010), 54.81 km2 (2015) and 54.87 km2 (2020). However, the error of the glacier area should be 147.11 km2 (2000), 146.99 km2 (2005), 146.38 km2 (2010), 73.34 km2 (2015) and 73.39 km2 (2020) with the consideration of the length of the ridgelines. We used the first calculation system in the manuscript, that is, the length of ridgeline wasn’t taken into account.
Related researches:
Sun, M.P.; Liu, S.Y.; Yao, X.J.; Guo, W.Q.; Xu J.L. Glacier changes in the Qilian Mountains in the past half-century: Based on the revised First and Second Chinese Glacier Inventory. J. Geogr. Sci., 2018, 28, 206–220. https://doi.org/10.1007/s11442-018-1468-y.
Duan, H.Y.; Yao, X.J.; Liu, S.Y.; Gao, Y.P.; Qi, M.M.; Liu, J.; Li, X.F. Glacier change in the Tanggula Mountains, Tibetan Plateau, in 1969-2015. J. Moun. Sci., 2019, 16, 2662-2678. https://doi.org/10.1007/s11629-018-5011-5.
Liu, J., Yao, X., Liu, S.Y.; Guo, W.Q.; Xu J.L. Glacial changes in the Gangdisê Mountains from 1970 to 2016. J. Geogr. Sci., 2020, 30, 131–144. https://doi.org/10.1007/s11442-020-1719-6.
Zhou, S.; Yao, X.; Zhang, D.; Zhang, Y.; Liu, S.; Min, Y. Remote sensing Monitoring of Advancing and Surging Glaciers in the Tien Shan, 1990–2019. Remote Sens., 2021, 13, 1973. https://doi.org/10.3390/rs13101973.
Zhang, Y.; Yao, X.J.; Zhou, S.G.; Zhang, D.H. Glacier changes in the Sanjiangyuan Nature Reserve of China during 2000–2018. J. Geogr. Sci., 2022, 32, 259–279. https://doi.org/10.1007/s11442-022-1946-0.
3) At present, in the relevant researches of glacier change, we usually interpret the glaciation region based on remote sensing images to obtain the glacier contour data in different periods. The relevant parameters (area, length, ice storage, flow rate, etc.) are analyzed with geographical significance in order to understand the glacier change in the region and its response to climate change. Statistical methods are inevitably used in this process, but it is worth noting that we analyzed the factors that may have contributed to the interpretation error before we analyzed the specific changes.
The revised manuscript has not changed meaningfully in its scope from the original submission. Since most of the measured changes are not significant, the scope is not appropriate. I believe a paper about non-discernable change can make a meaningful contribution. In this study, the author's findings support previous research about the area, using newer methods and more data. That finding can be important. Moreover, the authors have identified some glaciers changing within a mostly static region. That is also interesting. However, the manuscript continues to focus on telling a story about small changes, and these changes are often not statistically significant.
It seems that there is an interesting story that can be told about the data collected for this study. That story, however, needs to address uncertainty adequately. I encourage the authors to think about what information is presented, considering the small changes for the region and data constraints.
Response:
Thanks to the reviewers for their comments.
Indeed, the glacial changes in this region are insignificant compared to its own stock, glacial changes are very small. However, it is worthwhile that the relevant research results of this paper have some value for the explanation of the glacier anomaly of "Karakoram-West Kunlun Pamir". In the revised draft, we have revised and supplemented the evaluation and analysis of the uncertainty of glacier area change, and recalculated and verified the relevant data. Please refer to the revised draft for specific modifications.
General Comment, Introduction:
The introduction is much clearer. I appreciate the changes made. It makes it easier to follow. Also, it is more engaging
General Common On Diction:
4) In many places, the word significant/significantly is used. I believe it's being used to indicate these values are larger than the other discussion/region-wide average. However, significance is usually used when discussing results that surpass a threshold of a statistical test. For example, A and B are statistically different (i.e., the p-value is smaller than some pre-determined threshold for significance (e.g., 0.05) using a statistical test about two populations, such as a t-test). If statistical tests were not used to compare values, then I would advise again saying a value is significantly different.
Response:
Thanks to the reviewers for their comments.
The error of glacier area has been modified and redescribed in the revised draft for specific modifications, while the significance test you mentioned cannot be calculated at present. In the document, we provided the proportion of error to the initial value to supplement this section.
General Comment, On Uncertainty:
5) The overall uncertainty <75 km2 is surprisingly small. The uncertainty in 2000 is ~2.4%, which is quite small in light of values found in many other studies: usually 5% to 10% (or more). Equation 2 has the mathematical form of how 1-sigma uncertainties are propagated. Are the uncertainties presented here the 1-sigma uncertainties? In that case, the 2-sigma uncertainty values are usually presented.
Response:
Thanks to the reviewers for their comments.
For the calculation and evaluation of the area error, we informed the algorithm and its results in the second part of our response. Relevant references are also marked.
General Comment, Section 4.1 (General Change):
6) The 23.83 km2 of glacier area loss from 2000 to 2020 needs to be viewed in light of uncertainties. Using Equation 2, I get an uncertainty for the change of 82.20 km2, which greatly exceeds the magnitude of the change. Repeating that for the other time intervals, all the area changes are smaller than the uncertainty interval. Thus, it is not appropriate to state that the glaciers have changed over the entire interval (or any subintervals) or that there are changes in the loss rate over different time intervals. Any measured changes cannot be shown to differ from zero. Also, these changes are quite small (<25 km2) when compared to the scale of the glaciated area (>2,900 km2).
A similar analysis of the glacier length changes shows that nearly all of them are smaller than the uncertainty. If uncertainty in the change of length is propagated the same way as it is for the area (i.e., using Eq. 2), then all length changes are smaller than the uncertainty for change (84.8 m).
Response:
Thanks to the reviewers for their comments.
We are sorry that we were unable to make relevant changes and further reply to this question. The reason is that the glacier changes are insignificant compared to the size of the glaciers themselves, the glacier changes were very small. However, it is worth noting that the relevant research results of this paper have some value for the explanation of the glacier anomaly of "Karakoram-West Kunlun Pamir". In Section 4.1, we need to note that according to previous research methods and relevant parameters, we have provided information on the number, area and length of glaciers in different time periods and their errors. However, for the area error you proposed, which is much larger than the area change, our explanation can only be limited by the distribution characteristics of glaciers in this region. In the analysis in Section 4.2, even though the area error of some glacier changes was modified, it is worth noting that in previous studies on glacier changes, the authors usually only provided error information of total glacier changes, as shown in Table 1 of this paper. However, the analysis of glacier area changes in area grade, elevation and orientation didn’t provide relevant errors, which is mainly limited by the calculation method of errors and variation errors, and the effect as shown in FIG. 4 in this paper will be obtained.
General Comment, Section 4.2 (Characteristics of Glacier Are Change).
7) In the second half of paragraph 1, the newly added text (in red) contradicts what is shown in Figure 4. There are no discernable changes in Figure 4. In Figure 4b through 4g, the uncertainty intervals intercept with zero. In Figure 4a, the bar heights for the two intervals overlap in their uncertainty interval. Thus, what is said in the text is inaccurate from what is shown in the figure.
Response:
Thanks to the reviewers for their comments.
It is worth noting that in previous studies related to glacier changes, the authors usually only provided error information of total glacier change, such as the information in Table 1 of this paper, while no relevant error was provided for the analysis of glacier area changes in area grade, elevation and orientation, which was mainly limited by the calculation methods of error and change error. However, it is not possible to directly use a ratio for error assessment as you provided in the 8th opinion, so the effect as shown in Figure 4 will be obtained after calculation. In the revised draft, we recalculated the error of glacier change for different area grades in Figure 4, and did not consider the influence of ridgelines in the calculation process. Please refer to the revised draft for the modification of this part.
8) When discussing Figure 5, uncertainty should also be addressed. While there is no nice way to add an error bar to these types of plots, the values should be interpreted in light of typical uncertainty values. The largest bars for change I see are just shy of 10 km2. That interval has more than 400 km2 of glacier area. Thus, these changes represent at most 2.5%. That bar might pass the threshold for uncertainty. However, most of the other bars for change do not. The text discussing changes by elevation should consider the 'significance' of these small changes.
Response:
Thanks to the reviewers for their comments.
We have added uncertainty (2.5%) to the graph according to the algorithm you provided. significance has been changed in the document.
General Comment, Section 5.1
9) I appreciate the clarification on the method for including the climate change rate (and Figure 12e,f). However, I still have a question. My understanding is that a five-year sliding window was used to calculate a change rate. Are only change rates where the slope is statistically significant shown in Figure 12e,f)? If the p-value for the linear model is greater than the threshold (say 0.05), then the slope should be zero. Looking at Figure 12a,b, I am struggling to recreate the slopes shown in Figure 12e,f (by eyeballing local trendlines over 5-year intervals). Please clarify. And if the significance of the climate slopes was not considered, please reconsider the method.
Response:
Thanks to the reviewers for their comments.
Based on the mechanism of temperature change affecting glacier area change, it should be noted that this method is scientific and feasible on a longer time scale. The problem with this part of the study, in your opinion, is that the time scale is so short that the rate of temperature change is, in your opinion, not tested for significance and therefore meaningless. Here we believed that the statistical aspects should be downplayed and we should focus on the rate of temperature change, even if it is a statistical analysis of existing data based on the 5-year sliding slope provided by our paper or the subtraction of the two years you mentioned in the first round of comments, divided by the average year. Time scale and significance analysis should not be the conditions restricting its use. In addition, this method is based on the ideas provided in the doctoral dissertation of Lanzhou University, China, and we should admit that it has certain scientific and reasonable.
Reference:
https://kns.cnki.net/kcms2/article/abstract?v=3uoqIhG8C447WN1SO36whLpCgh0R0Z-iTEMuTidDzndci_h58Y6ouUHOgj_LOmJ-MK8xVl61hvHp-LD3Cdh1MydoGdCkfbZn&uniplatform=NZKPT
10) A pressing concern is whether there are different glacier rates to try to explain climatically. In Table 2, uncertainty intervals must be added to the Change rate column. And if those uncertainty intervals mean the change rates are not distinguishable from one another (or from zero), then the analysis in Section 5.1 seems inappropriate. Without discernible changes (and changes discernable from zero), any connections identified between glacier changes and climate forcings are likely spurious.
Response:
Thanks to the reviewers for their comments.
For the question you raised here, we need to note that our data is scientific and reasonable within the error range, so the area change trend should be accepted on the basis of abandoning the error. Otherwise, we would not be able to draw any conclusions in this article because they do not conform to the error assessment system you provided or constructed. For the existence of error, we should put it in perspective. We should not blindly use error as an obstacle to analyze data and draw relevant conclusions.
11) Is there a reference to another paper that has used this approach for the time lag method? I am still not sure that this is an appropriate approach. Was a linear regression model with a lag used and a 14-year lag was found to be statistically significant and the lag with the best r2 value?
Response:
Thanks to the reviewers for their comments.
The idea of estimating the response time of glaciers to climate change based on the trend of temperature change and glacier area change originated from the doctoral dissertation of Lanzhou University, China. Here we provide a link to the thesis for your reading.
Reference:
https://kns.cnki.net/kcms2/article/abstract?v=3uoqIhG8C447WN1SO36whLpCgh0R0Z-iTEMuTidDzndci_h58Y6ouUHOgj_LOmJ-MK8xVl61hvHp-LD3Cdh1MydoGdCkfbZn&uniplatform=NZKPT
General Comment, y-axis Ranges In Figure 14 & Figure 17:
12) I appreciate that Figures 4, 5, and 6 were reformatted to have the same y-axis range. The response to my comments indicated that was done for these two figures, but they still have different y-axis ranges. In one case (Figure 14f), the y-axis range is <0.4 km2, which makes for impressive-looking glacier length fluctuations. However, this glacier is almost 200 km2 large, making the fluctuations probably not significant.
Response:
Thanks to the reviewers for their comments.
The drawing has been modified according to your requirements, please see the revised draft for details.
Specific Comments:
13) Line 61: The change from Reference 21 is expressed as a percent and not as a rate (% a-1). I am unsure if it's a typo or if the findings from Reference 21 are the total change given as a percent. If it's the latter, then convert the findings from Reference 21 into a rate by dividing it by the time interval of their study.
Response:
Thanks to the reviewers for their comments.
According to your requirements, we have calculated the relative change rate of the data. Please refer to:
The findings indicate that the rates of change in glacier area within the West Kunlun Main Peak Area (WKMPA) during the periods of 1970s-1990, 1990-2010s, 1970s-2016 and 2010-2017 was -0.16±0.10%·a-1 [20], -0.01±0.32%·a-1 [20], -0.07±0.10%·a-1 [17] and 0.16%·a-1 [21].
14) Figure 1: The data source for the glacier polygons should be included in the caption. I assume this is from an earlier (and maybe larger regional) study. If so, cite it in the figure.
Response:
Thanks to the reviewers for their comments. We have added relevant literature in the manuscript for reference:
[20] Bao, W.J.; Liu, S.Y.; Wei, J.F.; Guo, W.Q. Glacier Changes during the Past 40 Years in the West Kunlun Shan. Journal of Mountain Science, 2015, 12, 344-357. https://doi.org/10.1007/s11629-014-3220-0.
15) Line 134-146: 'The range of threshold value ranged from 1.8 to 2.0' --> 'The threshold values ranged from 1.8 to 2.0'.
Response:
Thanks to the reviewers for their comments.
The section has been revised and we have adjusted it to “The threshold values ranged from 1.8 to 2.0”.
16) Figure 3: This is a great figure, but I didn't realize immediately that it is for two different methods. White space should be added between the glacier delineation and length extraction methods. Even better could be to break it up into (a) and (b) parts.
Response:
Thanks to the reviewers for their comments.
The section has been revised.
17) Line 173: 'large' --> 'larger'
Response:
Thanks to the reviewers for their comments.
The section has been revised and we have adjusted it to “To reduce processing errors, we used automatic methods of glacier classification, which have been found to be more robust than manual digitization for ice bodies larger than 0.1 km2 [45].”
18) Line 196: Please clarify what is meant by 'number of glacier length'. Is that the actual glacier length measure (e.g., 600 m)? If so, n seems like a strange variable to use. Or is n the number of pixels included in the length (e.g., 20 pixels for a 600 m long glacier)?
Response:
Thanks to the reviewers for their comments.
We mentioned in the article that the length of the glacier we mentioned is the flow line in the glacier. Previous studies have shown that the extraction basis of flow lines in glaciers lies in different types of glaciers, including common single glacier, compound valley glacier and ice cap glacier. For a single glacier, the length of the flow line in the glacier is the length of the glacier. For complex valley glaciers and ice cap glaciers, we usually take the maximum length of the flow line in the glacier as the glacier length. So the number of glacier length here should be changed to the number of glacier centerlines.
19) Line 214 & Table 1: It should be noted that it is the total glacier area and the average glacier length.
Response:
Thanks to the reviewers for their comments.
The section has been revised and we have adjusted it to “Table 1 presents the number, total area and average length of glaciers in the WKMPA during 2000-2020.”
20) Line 252: Identify the statistical test(s)/analysis(es) used.
Response:
Thanks to the reviewers for their comments.
The idea of altitude gradient analysis of glacier area change is to combine all glacier data into an object similar to an ice sheet, and then calculate the glacier area within each altitude gradient. The processing process was completed in ArcGIS, which was mainly used to analyze the changes of glacier area at different altitude gradients.
21) Figure 8a: Error bars are needed for the relative change rates to determine which (if any) of the values are distinguishable from zero.
Response:
Thanks to the reviewers for their comments.
We cannot provide an error bar for the relative rate of change here, because the error of the relative rate of change cannot be calculated under the existing calculation system. We can see that the error of glacier length in each period of this paper is 60 m, which has no influence on the analysis of the result of glacier length change.
22) Line 336: ‘terminal’ --> ‘terminus’
Response:
Thanks to the reviewers for their comments.
The section has been revised and we have adjusted it to “The change in glacier area is not solely attributed to terminus changes, but also influenced by the erosion of glacial lateral moraines.”
23) Line 343: I don't think this text has been updated considering the new bins in Figure 10.
Figure 10: I appreciate the changes, but some information has been lost. A three-bin range makes sense for Figure 7. If that same framework is used for Figure 10, are all length changes between -0.5 and 0.5 % a-1 indistinguishable from no change? I feel like the green bin is maybe too 'wide'. Perhaps a change of -0.4 % a-1 is meaningful. But if that is the case, such a change would be lumped into the green bin, which seems to represent no change. Maybe a 5-bin system would be better. There could be a light red and dark red bin for slightly significant and very significant retreats, respectively. The same could be used for advances but light blue and dark blue. I am not sure what would be the cutoff between no change and a slightly significant retreat/advance.
Response:
Thanks to the reviewers for their comments.
Here, we take your suggestions and set them into 5-bin system. Because the relative change rate of glacier length is -0.11% a-1, we set intervals at 0.2 intervals for analysis. Please refer to the revised draft for details.
24) Line 454 - 460 + Figure 16:
From where did the pre-2000 glacier data come? Were these findings from another study? If so, that study should be cited. Or, is this additional analysis that used earlier Landsat scenes than were used elsewhere in the study? If it's the latter case, then the earlier Landsat data must be identified and presented in the Methods.
Response:
Thanks to the reviewers for their comments.
In this part, we analyzed the yearly changes of West Kunlun glaciers before and after the split. The data came from the study in this paper, and the remote sensing image sources involved have been supplemented in the manuscript. Please refer to the revised draft for details.
Author Response File: Author Response.docx
Round 3
Reviewer 3 Report
I appreciate the time and effort the authors took to respond to my comments. However, we appear to be at an impasse. In response to my fundamental concern about the size of the changes relative to the size of uncertainty in the measurements, the authors seem unwilling or unable to account for uncertainty. Without doing so, the interpretations and presentation of the data in this paper are not accurate. In some cases, the presentation seems misleading. I cannot recommend this article for publication without significantly reworking the paper. At a minimum, it must accurately present what changes are discernable within the limitations of the data.
In a final appeal to the authors, I want to reiterate a point I made in the earlier reviews:
Over the 20 years, the area loss is 24.83 km^2 for a region with >2,900 km^2 of ice cover. That is a loss of <0.1% of the glacier area. Given that typical remote sensing uncertainties for glacier areas are usually well over 5% and closer to 10% or 15%, the region-wide changes in this study cannot be discerned. Also, using the author's estimate for the uncertainty of >50 km^2 to >100 km^2, I cannot conclude that there are [discernable] glacier changes in the region.
I have experienced frustrations like this in my own research. There have been times that I have built a glacier dataset from Landsat images that ends up not having discernable glacier-wide changes. Often, I can see the termini move a few pixels upslope in many valleys. But the uncertainty exceeds the retreat. It's frustrating. However, a richer story could be buried in the data that has not yet been discovered through the first set of analyses!
In writing this manuscript, significant time was likely spent collecting, analyzing, and presenting the data. I would like the see those efforts rewarded through a publication. However, many of the statements about the data in this manuscript are not accurate. Moreover, publishing a paper purporting changes that are likely not real could hamper research in the region as people will have an incorrect understanding of its glacial behavior.
Although the data indicate only minor (and mostly not resolvable) changes in the region, there is likely still a paper (or papers) that could be published from the dataset. At the end of this section, I suggest three directions for revision that the authors may want to consider.
A few remaining points for consideration/concern:
Point 1: I continue to have issues with the notion that there is a region-wide glacier response time (Discussion Section 5.1). I appreciate that the authors added a reference – Reference 17. However, reading Wang et al. (2018), I do not see a similar approach being used to estimate the response time. Wang et al. (2018) note that there is a lag between a climate forcing and the glacial area/length response. However, I don't see in Wang et al. (2017) any attempt to estimate that lag time.
For details on how to estimate glacier response time to a forcing, I would recommend the following article by Gerard Roe:
Roe, G.H. and Baker, M.B., 2014. Glacier response to climate perturbations: An accurate linear geometric model. Journal of Glaciology, 60(222), pp.670-684.
A glacier's response time depends on many factors, including its size, bedrock slope, mass balance, etc. I would expect individual glaciers in the region to respond at different rates to the same forcing. Thus, an approach that finds a single lag time value between a region's climate forcing and glacier response time seems fraught.
Point 2: Regarding the rates of climate change in Figure 12a,b, I thank the authors for clarifying the methods. However, knowing more about the methods, I have a major concern. If I understand correctly, the rates are calculated by taking the climate variable value for the first and last year of the 5-year interval, differencing them, and then dividing by 5. If that is correct, then that is not an accurate method for estimating the 5-year climate trend. Differencing the first and last years' values makes the trends extremely susceptible to outliers. For variables where the interannual variability is much larger than any long-term trends – such as temperature and precipitation – calculating the 5-year trend by differencing the first and last year would be fraught.
A better approach for finding a 5-year trend would be to use a 5-year sliding window and perform a linear regression on each [sequential] 5-year interval. The trendline's slope would be the rate of change in that climate variable. That rate would reflect the values of all five data points in the 5-year interval – not just the first and last data points. An added bonus of this approach is that the authors could determine the significance of the slope values. Most regression packages can perform statistical tests. It would only be appropriate to include climate trends that meet some significance threshold. Since climate data are often noisy, I have seen researchers relax their p-value threshold from 0.05 to 0.1. Even with a more relaxed p-value (0.1 instead of 0.05), I suspect many slopes would not pass a significance test.
Point 3: In Table 2 and Section 5.1, I have to disagree that uncertainty can be excluded. Using the data from Table 1, I have calculated the change rates with uncertainty. They are as follows: -2.79 +/- 31.11 km^2 a^-1, -0.33 +/- 31.03 km^2 a^-1, 0.15 +/- 24.45 km^2 a^-1, -2.00 +/- 15.51 km^2 a^-1. The uncertainty is substantially (an order of magnitude) larger than the central values.
Failing to include uncertainty in the 5-year retreat rates misleads the reader. When including the uncertainties, it is clear that we cannot say any of the change rates are different from one another (or different from zero). It is also worth noting that even the differences in central values are minuscule. The largest difference in central-value retreat rates is <3 km^2 a^-1 for a region with more than 2,9000 km^2 of glacial area. Given the lack of discernable changes in the region (and the minute differences in the central values), the analyses in Section 5.1 seem inappropriate.
Some Thoughts On How One Might Revise To Address Issues Raised Above:
Approach 1: The paper could use the analyses conducted but faithfully present the data shown in the visuals made (with some slight medications). That would involve being explicit that most changes over the twenty years are not discernable. For example, the 24.83 km^2 loss over the 20 years represents a loss rate (over the 20-year period) of -0.04 +/- 0.21%. The central value and uncertainty must be included in the text. Without it, the readers would need to dig into the tables, fire up Excel (like I did) and crunch the numbers themselves. A change of 0.04% could be important, depending on the measurement accuracy/uncertainty. In this study, however, the uncertainty is a factor of 5 larger than the central value.
In Approach 1, some of the phrasings in the text would also need to be adjusted. For example, Line 218-219 states: '... it was found that the most significant changes in glacier area and length occurred between 2000 and 2005, with relative change rates of -0.09%·a^-1 and -0.23%·a^-1, respectively.' That statement would need to be revised as something like the following: 'The largest change in glacier area and length occurred between 2000 and 2005, with relative change rates of -0.09 +/- 1.04 % a^-1 and …, respectively.' Removing the word significant and explicitly including the uncertainty accurately conveys the values and allows the reader to know their meaning.
Although there are no significant changes, these numbers are still interesting. They can be used when comparing this study's findings with other studies and by future studies when comparing their results to this study. Approach 1 is probably the least onerous type of revision. Even it will require substantive changes to the presentation of the data (especially in the text). In rephrasing how the data are presented, the authors may stumble upon previously unseen (or unrecognized) findings!
Approach 2: Another way to reframe this paper could be to change its focus/scope. This paper confirms the findings of previous studies that indicate that this region behaves very differently than the rest of the Tibetan Plateau. The focus could shift to being a confirmation study of this past work. Part of a confirmation paper could also highlight where there are discernable changes. Figure 7 and Figure 10 illustrate that there are some glaciers with discernable area and length changes for this region, which is otherwise relatively static. Highlighting the glaciers that are discernably changing would enrich a confirmation paper. Approach 2 would be a more substantial revision than Approach 1, but it would likely lead to a more interesting paper.
Approach 3: Yet another revision avenue could be to explore why [most] glaciers in this region are not changing by a discernable amount. And why any changes are much smaller than in other locations (in the Tibetan Plateau or globally). It seems like there is some warming in the region (I kind of see it in Figure 12c). Maybe moistening in the 2000s (I see a lot of upward-facing bars during that interval in Figure 12d) has compensated for the warming. And/or maybe other meteorological variables/processes compensate for the increased melting from warming. Or perhaps there is a negative mass balance but the response time is so long that the glaciers have yet to appreciably retreat. Approach 3 would be an extensive revision but may lead to interesting findings.
Given the scientific issues highlighted in my Comments and Suggestions for Authors, I did not include line-specific editing comments.
Overall the paper is easily readable. I would be happy to provide detailed Quality of English comments on a revised version.
Author Response
The authors have responded satisfactorily to the indicated issues.
Author Response File: Author Response.docx