The Normalized Difference Snow Index (NDSI) is vital in snow cover extent, snow cover fraction, and snow depth retrieving in case of using optical satellite observations. In this study, the authors developed an integrated cloud-free MODIS NDSI over China with the help of a spatio-temporal Adaptive fusion method. However, the following global and regional cloud-free gap-filled NDSI and snow cover extent dataset are excluded in this study, neither in the introduction section, nor in the cross-comparison section. Therefore, it is difficult for me to evaluate whether this dataset is uniqueness or usefulness.

The existing cloud-free NDSI dataset and snow cover extent datasets including:

• MODIS/Terra CGF Snow Cover Daily L3 Global 500m SIN Grid, Version 61 (https://doi.org/10.5067/MODIS/MOD10A1F.061), which provide a cloud-gap-filled daily MODIS NDSI dataset at 500m spatial resolution.
• The NIEER AVHRR snow cover extent product over China (https://essd.copernicus.org/articles/13/4711/2021/, https://doi.org/10.5194/essd-13-4711-2021), which provide a cloud-gap-filled daily AVHRR snow cover extent dataset over China.
• The daily MODIS 500m snow cover extent over China (http://data.casnw.net/portal/metadata/be3a4134-2e5c-467f-8a5e-b1c0ed6cc341, doi:10.12072/ncdc.I-SNOW.db0001.2020)
• Daily fractional snow cover dataset over High Asia during 2002 to 2018 (http://www.ncdc.ac.cn/portal/metadata/0e277d66-d89b-4e54-8a75-fe22fcc3adee, doi: 10.11922/sciencedb.457)

Although the study may content material worthy of publication, the paper in its current version needs major revision and resubmission to meet the level expected of ESSD, for the following reasons.

1. Literature review in Introduction disregards former studies on NDSI retrievals from satellite data, including both datasets retrieval methods. Thus, the contribution of the current study in accordance to existing knowledge and methods is not clear. Please add the above listed cloud-free NDSI dataset and snow cover extent datasets in the introduction section.
2. The lack of innovation in accordance to existing knowledge. Please add the comparison between NDSI dataset in present study and MODIS/Terra CGF Snow Cover Daily L3 Global 500m SIN Grid, Version 61, both in fusion method and results.
3. The lack of depth in the result analysis that makes the study inconclusive. Please emphasize the unique contributions in the present study in the comparison with the above listed cloud-free NDSI dataset and snow cover extent datasets over China.

Comments to authors

The manuscript titled “STAR NDSI collection: A cloud-free MODIS NDSI dataset (2001–2020) for China” estimates cloud-free snow data for China. The authors use Spatio-Temporal Adaptive fusion method with erroR correction (STAR) to derive snow cover. Cloud cover is the main obstacle in passive remote sensing snow monitoring and is important to overcome. The study is important but there are few major issues in the present form which needs to be addressed.

Major comments

1. The authors use combined Terra and Aqua MODIS data in this manuscript. They combine the data first and then use STAR method consisting of spatio-temporal adaptive fusion (STAF) and error correction (EC). Combining Terra and Aqua this way potentially overestimates snow (Muhammad and Thapa, 2020, 2021). The authors are suggested to either revise the TAC or explain the potential uncertainty.
2. The authors missed to share the code to generate STAR NDSI dataset. It is incomplete without sharing the code. The code is also required to evaluate the methodology as well.
3. The C6 snow is in NDSI ranging between 0 and 100. It is not explained how the authors reconstructed the snow data. It is a challenge to improve the data on how to replace the cloudy pixel, so it is significant to understand the way the value is replaced.
4. The authors indicate they have derived data between 2000 and 2020. The Aqua data is available from July 2002, the authors should clearly mention the observed period. As the data is combined Terra and Aqua, therefore, it should be between 2002 and 2020 not starting from the year 2000.
5. One of the major issues is the remaining overestimation. The authors have to consider the existence of overestimation mainly due to the larger solar zenith angle. It is, therefore, necessary to estimate the overestimation in the combined Terra and Aqua as in the combined product the uncertainty increases.

Minor comments

Line 45-65: The authors missed to point out one of the most recent cloud-free 8-day (Muhammad and Thapa, 2020 - https://doi.org/10.5194/essd-12-345-2020) and daily (Muhammad and Thapa 2021 - https://doi.org/10.5194/essd-13-767-2021) snow data, combining Terra and Aqua satellites data, reducing up to 50% of uncertainty. These datasets uniquely combine Terra and Aqua, to avoid overestimation after temporal and spatial filters are applied to individual products for clouds removal. The authors are advised to add these important papers.

MODIS is onboard on Terra and Aqua satellites, the authors are advised to clearly mention which constellation they use in e.g. to clearly mention in “A daily spatio-temporal continuous MODIS C6 NDSI dataset with a spatial resolution of 500 m for China (Fig. 1) from 2001 to 2020 is generated for the first time.”