This manuscripte produced a new MODIS snow cover product over China. It  includes Terra SCE and Aqua SCE datasets, as well as a cloud-gap-filled SCE dataset. Validation against with in situ snow depth measurements, these products show obvious improvements than standard MODIS SCE products. The produced snow cover extent product could be a significant dataset for climate change research.

Despite of its significance, several issues still need to be resolved before a possible publication to HESS. 1) OLI snow product is very important for optimal NDSI thresholds, and how do you obtain them? You should describe it more sufficiently; 2) Organization of the paper should be improved, especially section 3. I am confused by the first two paragraphs of section 3. Why you introduce how to obtain the samples here? I think maybe putting them in section 2 is better. More importantly, you should show a flowchart first in section 3.1. Otherwise, it is difficult for readers to understand how you will produce SCE from MOD09GA or MYD09GA. There is a same problem for section 3.2. I would suggest you reorganize this section. 3) Some minor errors that may need to be modified are listed as follows.  

Minor comments and suggestions:

1. Line 48-55: People who are unfamiliar with MODIS snow products may be difficult to understand your introduction on these products. I would suggest you give a clearer description.
2. Section 2.1 line 88-93: You should introduce the MOD09GA, MYD09GA and MCD12Q1 concisely.
3. Section 2.2: See the first suggestion. A simple introduction on the OLI snow maps is definitely needed.
4. Section 3. 3.1.1 line 130-138: Preliminarily screening, it is repeat for the last two graphs.
5. Section 3.1.2 line 154: Optimized NDSI thresholds, "However, as expected, only using the NDIS criterion seems not accurate enough to discriminate snows over those forest land-cover types, except the "Evergreen Needleleaf Forest" (due to its sparse distributions in China)." Change "NDIS" to "NDSI"!
6. Section 3.1.4: Postprocessing based on surface temperature and DEM, how to determine the threshold of surface temperature screen?
7. On Figure 2, please provide the full name of Figure 2.
8. Section 3.3.2, line 230: “For the aggregated SCE data”-> “For the aggregated SCE”.
9. Section 4.1: Confuse matrix is a commonly-used tool to evaluate the products relevant to classes. It seems this section is unnecessary.
10. Section 4.3, line 268-270: This may be attributed to different snow/non-snow number distributions in nature among different years, and varying sample numbers caused by different ground measurements available in different years. I cannot understand “different snow/non-snow number distributions in nature among different years”.
11. Section 5.2, line 325-335: for the two examples, are they all covered by forest?
12. Line 355: During our validations or comparisons, we found this phenomenon is somewhat common in the edges of snow-cover areas and the forest areas of Northeast China. Very awkward sentence. Please consider revising it.
13. Section 6, line 345: finally, a totally cloud-free SCE is mapped through replacing the residual gaps with auxiliary passive microwave snow-depth data. Is “finally the residual gaps are all filled according to the implication given by a auxiliary passive microwave snow-depth dataset” better?
14. Section 6, line 350: “by a series of processes filling cloud-induced gaps”. It seems wordy here because you just mention them in the above paragraph!

This study proposes a new MODIS snow-cover-extent product over China. The optimal NDSI thresholds varying with land cover types are extracted, the NDVI-NDFSI decision rules specific for snow discrimination in forest areas are optimized, and an HMRF gap-filling technique, which can simultaneously assimilate temporally and spatially neighbouring information, is imported. The need for such an approach is well justified and the authors cite ample relevant literature. The study provides examples demonstrating the successful performance of the method. The paper is basically well-written and presented.

There are a few important and minor comments/mistakes that are listed below and should be taken into account.

Line 38, “on the other hand”, where is “on the one hand”.

Line 53, “the MODIS band 4 (0.55 μm) and band 6 (1.6 μm) reflectance” should be “the reflectance of MODIS band 4 (0.55 μm) and band 6 (1.6 μm)”.

Line 54, “distinguish snow cover or not” should delete “or not”.

Line 58, “research” should be “studies”.

Line 78, the use of GEE is not motivated from the above text.

Line 80, “surface cover” should be “land cover”.

Line 88, “chiefly” should be deleted.

Line 88-91, all of the definite articles before the acronyms should be deleted. All of the “us” also need to be deleted.

Line 102, only the snow-depth product has linkage?

Line 106, “too” should be deleted.

Line 108-115, why was snow-depth data used to validate SCE product? It is not a snow depth product. This need explanation in the text.

Line 118 and 234, “paper” should be changed to “study”.

Line 119, “Sect.” should be “Sec.” or “Section”.

Line 154, “NDIS” should be “NDSI”.

Line 167, “≥” should be “greater than”.

Line 187-191, I suggest to move Subsection 3.2 to the front of Subsection 3.1, and to combine Subsection 3.2 and 3.1. The authors should give a general introduction of the method first, and then present the details. There is the same problem in Subsection 3.4 and 3.3.

Line 281, “this section” should be changed to “we”.

Table 3, 4, 5, and 6, OE is just one minus PA and CE is just one minus UA, thus OE and CE are not need to show if you use PA and UA, and PA and UA are not need to show if you use OE and CE.

Line 338 and 340, I suggest to delete “Under the support of several national programs of China” and “Toward this”.

Line 345, “and finally” should be change to “finally”.

Figure 1, 2, 3, 4, and 6 use “non-snow”, but “snow-free” is used in the text and Figure 7.

Figure 3, “Number density colorbar” should be “Number density”.

The parentheses in the caption of the figures should be deleted.

This manuscript develops a novel method of creating seamless snow cover extent (SCE) data for China. The novelty of this study is significant in that there is currently no other dataset of SCE with such a long time series and quality (e.g., a full spatiotemporal continuity at 500m and daily resolutions, sufficient evaluations, and accuracy) that can be comparable with here developed. The improvement of performance in forested areas is also impressive. I have some minor concerns as follows.

• Landsat-8 OLI data are used as training data in this study, however, how were they processed as reference SCE data? By manually vectorizing or also using an NDSI-like method? This process of course has uncertainties that should be mentioned.
• Why were two groups of Landsat-8 snow maps used for training Terra and Aqua separately? I do not think the overpass time is a good reason for explaining this because most Landsat-8 data are in the morning which is closer to Terra. This could also be an uncertainty for training Aqua data.
• I am a bit confused by Table 2 because there are several duplicate land cover types in the first column and some types with multiple rows. It can be improved to be clearer.
• In section 2.4, I recommend adding more explanations for the use of station observations. I can understand that the station data were used as they were totally free of the effects of cloud blockage. However, for some readers, this could be confusing when Landsat8 snow maps were already used as reference.
• For section 3.1.4, the use of ERA5 LST could be a (possibly not big) problem or not optimal. The MODIS standard product uses MODIS LST with a high resolution of 1km in the aid for temperature screening, however, ERA5 only has a coarse resolution of 0.25° which is much larger than the pixel size of MODIS SCE data.
• In Figure 7, the microwave-based snow depth data are suggested to be additionally plotted against other data. It can be seen in Figure 7 that there are large data gaps in both Terra and Aqua snow map on that day and the microwave-based snow depth data could provide valuable reference data of snow in those areas.
• The English writing has a large potential to be improved.

Typo errors:

Lines 65-66: please add a reference

Lines 133-138: the two paragraphs could be combined to reduce duplications

Line 154: NDIS should be NDSI

Line 213: gas should be gap?

Line 300. Terra? I guess you’d like to say Aqua?