Dear Sam,

I read your reply and the revised version. Most of the comments were well addressed or I agree with the argues in case you did not follow. I did therefore not send the manuscript for further review, but have the following minor comments:

General comments:
In a strict sense you are using a digital surface model (DSM) of the glacier (surface). This term should at least be mentioned in the beginning.
More and more higher resolution DEMs are becoming available, but for many parts of the Earth there is no DEM with a resolution of 5m or better available and I would also not (yet) see these high resolution as “conventional”. If available for free, these DEMs with relatively high resolutions provide often a merge of several scenes and only one snapshot in time. However, cliffs are changing. In addition, they can contain significant errors as the data is processed for large areas automatically. This should be clarified and the term “conventional” avoided for the DEMs needed. You may also mention the new HMA DEM with a resolution of 8m (Shean, 2017) and also the TanDEM-X DEM with 12m resolution.

Specific comments:
P4L7: <1 m should be written in brackets
P4L13: You cannot start a sentence with “<1”. Reformulate
P4L18: What is .17ex~5 m resolution? Should be clarified. In addition, you should clarify here that one does not need a DEM with a resolution of 5m, but that the method works also for lower resolution DEMs.
L20: But they should also not be too different in time. Debris-cover is changing.
P5L4f: Be specific; which sensor with which resolution did you use? Mention the advantage of S2 (10m res. multispectral vs. 30m (15m pan).
L12: “Structure from Motion” is a method not data. Clarify to which data you refer to.
L15: I agree but you do not need high absolute height accuracy with respect to the geoid.
L17ff: Provide a reference or URL for the Arctic DEM and mention here also the new HMA DEM and other suitable DEM sources which might be available for free for scientific purposes (such as the TanDEM-X DEM)
P7L12 “was” identified. Write here which sensor was used (OLI?)
P8L24: “Eq. ??” ?
P15L12f. (and also elsewhere): I do not like sentences like” Table 3 shows…” as this is a bit waste of space as Table/Figure captions are often similar. Write the most important finding and then refer to the table in brackets. However, this is a matter of style and my personal view, so feel free to leave these sentences as they are.
P17L10 “2x” -> two times
L12ff: See my comment P15L12f, but more important: I find the comparison of the results based on different resolutions very relevant and interesting. This should be extended a bit more. If you could compare also to freely available DEMs for the same glacier(s) (HMA DEM vs. TanDEM-X, SRTM1, National DEMs etc. depending on coverage) for the same glacier to show the differences would from my point of view clearly increase the impact of the paper.
P18L3 “geomorphic”
P20L13: “Kääb”
Table and Figure captions: The captions are partly very long. It is again a matter of style but from my point of view long captions distract from reading the main text and with clear Figures (incl. legends) and Tables (incl. clear headings for the columns) long captions are not needed. Baseline information from the main text is also not needed (e.g. DEM resolution for this study was 5m).
Tables 2 and 3 could be shown in one. There is enough space as the letters of the tables are even large than of the text.
Figure 1: I find it hard to read. I would clearly prefer a usual map projection (maybe centred in eastern China or Japan). I cannot identify the location of the both glaciers on the map. I find also the oblique view not well done and the black lines hard to identify. I would use nadir view and do not cut most of the upper glacier part. The sentence “In the oblique inset maps, orange lines are the glacier extent and the black polygons define the spatial domains used in this study” can fully be avoided when including a legend.
Figure 2: The letters are too large – waste of space and visually not really nice. Show the area of the subsets in Fig. 1.
Figure 3: What is the source of the optical data used?
Fig. 4: In line with Rev. 1 I’ also show Le on the other side.
Figure 5: In line with the reviewer I ask to flip the colour coding for the temperature. It needs to be consistent with the other figures and cold should be blue. In case you want a similar look you may flip the ice cliff probability.

Do not hesitate to ask in case you have any questions.

Best regards,
Tobias

Reference:
Shean, D.E., 2017. High Mountain Asia 8-meter DEM mosaics derived from optical imagery, version 1. Doi:10.5067/KXOVQ9L172S2.

Dear Sam,
dear Francesca,

Thank you for carefully improving the manuscript. I am happy to accept it now as is.

Thank you for contributing to TC and best regards,

Tobias - Editor TC