Dear Wei Ji and Huw Josepth,

You have had two globally positive reviews for your manuscript. Thanks you for your careful reply to both of them. I now strongly encourage you to submit a revised version of your manuscript that account for all these constructive comments. Based on this new version of your manuscript I will take a final decision regarding its acceptance to The Cryosphere.

Regards,
Olivier Gagliardini

Dear Wei Ji and Huw Josepth,

Thanks for your careful reply to both reviewers and the improved version of your manuscript. Based on that, I am happy to accept your paper for publication in The Cryosphere after some minor revisions have been done (see below).

Thanks for having choosing The Cryosphere to publish this nice and innovative piece of work!

Regards,
Olivier Gagliardini

Minor revisions :

- section 3.1 (or elsewhere?): would be nice to quantify the dataset used for the training relative to the whole dataset (number of pixels / total number of pixels or surface area relative to the total area).

- below Eq. (3): are theta and eta vector or scalar parameters?

- line 160: avoid starting a new sentence by a symbol (here you can continue the sentence with "and, ")

- line 176 (and check all along the manuscript): see Fig. 1

- line 243-244: I don't understand why the resampling is needed for a 1d comparison? Why not using the native dataset?

- line 329: see Tables 1 and 2

- line 379: more than independent, I understand your method is complementary as it rely on an initial BedMachine or BEDMAP DEM, correct? As far as I understand, it would not work without an input from a large scale DEM like BedMachine or BEDMAP, which would be the requirement for an independent method?

Dear Wei Ji and Huw Josepth,

Thanks for these last corrections and thanks again for having choosing The Cryosphere to publish this nice and innovative piece of work!

Regards,
Olivier