|The revised version of the article entitled ‘InSAR time series analysis of seasonal surface displacement dynamics on the Tibetan Plateau’ from Reinosch et al. have been subject to an intensive work of corrections. I would like to thank the authors for having assessed carefully the major comments from my previous review. The manuscript is much clearer, explains better the methods and presents a more comprehensive discussion of the results. I believe the manuscript is suitable for publication subject to minor revisions (see suggestions thereafter).|
1) Good that the temporal baselines are now documented (l.183-184). But it rises a new question: why the range is different for the two basins: 12-60 (Niyaqu) and 12-96 days (Quqaqie). Good to quickly explain it. It is potentially a problem to compare the results cause the temporal baseline is related to the detection capability. Decorrelation can occur from 10.6 cm/yr for Quqaqie and 17 cm/yr for Niyaqu (and half of this for aliasing). You partly discuss it in 6.3 but good to also mention the max. velocities in ’Methods’ and think about how it may have affected the comparison between the basins.
2) The description of the different models is much easier to understand, but the table could still be improved to my opinion. Here some suggestions: in ’Purpose’: for ex. LVM: Multi-annual subsidence, sediment accumulation and permafrost creep. HSM: Seasonal heave-subsidence cycle from AL freezing and thawing. SSM: Seasonally accelerating slopes processes (as linear creep is included in LVM). In ’related geomorphological processes’: LVM: add long-term subsidence and sediment accumulation, as you are also including <10 degrees areas. In ’associated landforms’: In HSM: I guess this model is not only considering hummocks - maybe good to be a it more general. In SSM: debris mantle slopes, solifluction lobes, rockslides.
3) The discussion has been much improved. I think there are maybe still some elements missing related to the difference between ascending/descending results and the two basins: first, the temporal baseline is different (see comment 1). It can partly explain the difference in coherence and it means that you may underestimate more high velocities in Qugaqie than in Niyaqu. Second, the time interval in Qugaqie is different for the 2 geometries (starting in June for ascending and November for descending). You have one more summer season in ascending geometry in Qugaqie. Third, the spatial distribution of the DMS difference: looking at S12/S13, it seems that the shift between ascending/descending is not randomly distributed, it has a spatial pattern. I have not thought much more about the interpretation of these 3 elements and how this could influence your results, but I believe it should be considered.
l. 33: Is ‘collapsing moraines’ the good terminology here? I believe this would refer to quite sudden and catastrophic events. Maybe just ‘frozen moraines’? Ice-cored, push-moraine or just a frozen morainic deposit may be affected by permafrost creep or thaw subsidence, without necessarily collapsing.
l. 34-35: As you show now on Fig.6, even these processes have some seasonal patterns. Consider rephrasing to ‘...show little seasonal variation but a linear pattern indicating that their displacement is predominantly gravity-driven.”
l. 39: Formally the active layer is not part of the permafrost (as it is not following the definition written at l. 38-19): it is the layer above the permafrost that freezes and thaws seasonally.
l. 40-42: I would rephrase to ‘This causes frost heave and thaw subsidence of wet ground on the order of centimeter...’ Doesn’t need to be saturated to heave/subside (as you acknowledge at the next line by written that the amplitude depends on the water content). Check also the English rule, I would also tend to say that it is better to use ‘heave/subsidence’ instead of ‘heaving/subsiding’.
l. 47-48: As the first paragraph is a general introduction, consider to move the sentence about TP in the next paragraph.
l. 65: What do you include in the terminology ‘slope instability’? It is unclear in this sentence and looks unbalance to list ‘...such as slope instabilities and the creeping of rock glaciers’
l. 66: here before to mention laser scanners and subsurface data, it could be good to mention in-situ surface methods, such as DGPS (more used historically and worldwide than lidar): for ex. ‘… can be monitored by collection of in-situ surface or subsurface data (ref), or terrestrial remote sensing techniques, such as laser scanning (ref)’.
l. 75: Here the focus should be on the temporal coverage, as you just wrote about cloud cover for optical time series. Maybe rephrase to ‘… make the systematic/continuous detection of these displacements possible’.
l. 94: High altitude itself is not a problem, the problem occurs when there is a large variation of altitude over the investigated area (the stratified atmospheric problem come from the difference of delay between valleys and mountain tops)
l. 84-96: I wonder if some of this shouldn’t be moved to ‘Methods’.
2. Study Area:
l. 110: NAMORS, Fig. 1A
Fig. 1: You probably don’t need to have information about data source both on the figure and in the legend. I would remove the small texts in the figure: hard to read, redundant and so a bit useless.
l. 131: ’prime study site for surface displacement using InSAR technology’ or ’prime study site for surface displacement related to periglacial processes using InSAR technology’
l. 145: Something wrong with this part of the sentence: ’as they represent different levels of glacial impact and the predominant landscapes and their related surface processes at Nam Co’
l. 149: Do you mean that the remaining area is not permafrost but seasonally frozen ground? If yes, just write it this way. I wouldn’t say that periglacial processes are limited to the higher parts of the sub-catchment. Periglacial is a quite vague terminology that does not necessarily mean permafrost. It includes processes in seasonally frozen ground.
l. 173-174: Just as a comment: 4x1 is quite little mutli-looking and may lead to a bias in coherence estimation, i.e that low coherence are overestimated (see e.g Bamler & Hartl, 1998)
l. 175-180: All this explanation about ‘poor interferograms’ is a bit long and vague. Do you mean low coherence? Maybe possible to just compress this information and say sth like: ‘we started our times series in xx 2014 for xx basin and xx 2015 for xx basin due to low coherence in earlier acquisitions.’
l. 183-184: See main comment 1
l. 203-205: ‘...with a short temporal baseline, meaning the time time between the acquisitions was short’: SBAS means Small BAseline Subset (both spatial and temporal) but what is ‘small’ is relative to the process under study. Maybe write instead: ‘The SBAS method generates interferograms between SAR acquisitions with a temporal baseline under a chosen threshold, and stacks them to estimate displacement and velocity over a longer time period.’
l. 119-221: Weird that these sentences come here, before the part about topographic phase removal and trend correction. Maybe move them at the really end of 4.1 section.
l. 224: ’we then estimated and...’
l. 233: ’ We therefore performed a linear correction to remove this spatial trend from both ascending and descending datasets.’
l. 243: Not sure LOS and reference areas are comparable -> the LOS measurement is a projection of the real vector in the LOS direction, leading to a potential underestimation. It has nothing to do with the spatial relativity (to allow the conversion from cyclic phase to absolute displacement). I would just write ’InSAR displacement products are spatially relative to chosen reference points or areas.’
l. 244: Points or areas are both used (’areas’ in the title of 4.2, sometimes ’points’ in the text, ’points’ in supplementary). Maybe choose a similar terminology. I guess practically you are selecting pixels (=areas with the resolution of your datasets).
l. 263: Point 3) could be merged with point 1). As I understand it, it is actually the same reason. Point 4) = have a good coherence, it may be placed as first point.
l. 292-295: The areas with near linear rate are included in LVM, right? So it would be maybe more clear to not mix it again when speaking about SSM. Sth like ’The SSM focuses on slopes where sliding is accelerated from spring to autumn, to be differentiated from slopes in LVM where sliding takes place throughout the year at a near linear rate.’
Table 2: See main comment 2
l. 321: I don’t think ref. to Fig.3 is well located here.
l. 328-334: ’The larger the difference between the LOS vector and the vector representing the assumed motion direction, in this case downslope, the smaller and therefore stronger the downslope coefficient becomes’: It is not the most intuitive sentence ever. And in general the whole explanation about the coefficients is quite long and heavy to follow. Consider to rephrase and shorten. For ex. just add a map with the coefficients in supplementary and refer to it without too much details.
l. 351: I guess you mean ’...representing the seasonal signal...’, right?
l. 357-358: ’There is a lag time between the maximum air temperature and the DMS which has been documented to be 2 to 4 months on the TP (Li et al., 2015; Daout et al., 2017).’ This is already results/discussion.
l. 380-383: Try avoiding repetition. This is sth you already explained in Methods.
Table 3: Are the 3 numbers in the column ‘unstable’ corresponding to uplift/subside/move horizontally? Not completely obvious.
Figure 3: In legend, what do you mean by ‘relevant parts’. Sounds a bit unnecessarily wording.
Figure 3: As this combines two different projection according to the slope angle, it would be nice to add in supplementary a slope map or a map with the classification </> 10 degrees to see where each projection has been applied.
l. 444: If you want to repeat the meaning of the acronym (not necessarily a bad idea), better to do it at first appearance in the section (l. 441).
l. 446: ‘The median shift error’: not clear what it is and how you calculated it.
l. 464: Is this second station closer to the Niyaqgu basin? If yes, this just explains the lag: the mean in temperature is later in the second basin and the NAMORS station is thus not fully representative for this. You could just state it a bit more clearly. You are going further with this question in the discussion, but should anyway try to provide a quite standalone formulation in this section. Either you don’t mention it yet and refer to the discussion section for more info, or you give a bit more explanations.
l. 478 and Figures 6-7: Would you say one cause of the acceleration is the water input from the monsoon? If yes, it is maybe misleading to call the patterns ‘freeze-thaw-driven’.
l. 505-511: If it is not too much work, it would be nice to include in supplementary, one example for each identified category (orthophoto or field pictures).
l. 511: Same comment than previously about ‘collapsing moraine’. Not sure it is the good terminology.
l. 528-537: The start of this paragraph is a repetition. Not necessary I would say. And the end does not fit here. It should probably be in Methods.
l. 555: Sth wrong with the sentence ‘by providing more water during the freezing process’. Rephrase.
l. 559-660: ‘but it indicates that the maximum air temperature is too similar to that of NAMORS to explain the short lag time.’ It is more than a week of difference. I agree with the following discussion (shorter thawing season, thinner active layer, etc.) but before that, it could be stated a bit more clearly that using NAMORS station, the lag is most likely underestimated.
l.586-625: See main comment 3.
l. 621: ‘Heave-subsidence’ instead of ‘heaving-subsiding’
l. 622: ‘...where R2 >0.9 and rises to 29 where R2 <0.6’. How did you calculate sqrt R (in general, there are several locations where you present error estimations and it is not clear how it has been found out. 2-3 lines could be added in ‘Methods’).
l. 646: ‘...coherence values in in the center...’: twice ‘in’
l. 660-694: Shouldn’t the part about linear patterns be moved in the previous section (6.3)?
l. 675-677: ‘It is therefore possible that fast linear velocity is an indicator for ice-driven landforms. This disagrees with some other studies observing strong seasonal variations in the velocities of rock glaciers’ I don’t really understand the sentences, especially with the explanations that follow. 36% linearly creeping are rock glaciers does not make linear velocity an indicator for ice-driven landforms and does not disagree with observations that rock glaciers can in some cases experience strong seasonal variations.
l. 696-697: The conclusion is supposed to be quite standalone, so good to repeat quickly that it is based on Sentinel-1 InSAR results.
l. 704: maybe good to mention solifluction and rockslides separately. A rock slope instability can occur on slopes that are not much debris-covered (fractured bedrock), and as you mentioned previously can have a linear pattern.
l.709: ‘...indicating that they are not related to freeze-thaw processes.’ Rather say that it is gravity-driven. A rock glacier has also something to do with freeze-thaw processes...
S1: Add in legend that yellow acquisitions correspond to the master scenes for the co-registration.