This paper proposes a long-term AVHRR snow cover extent product from 1981 until 2019 over China. The product has the spatial resolution of 5-km and the daily temporal resolution, and is a completely gap-free product, which is produced through quality control, cloud detection, snow discrimination, and gap-filling. The validations based on ground measurement and Landsat-5 snow maps both demonstrate its higher accuracy than that of the JASMES AVHRR product. As a long-term record, the dataset will provide a valuable data source for analyzing the influence of climate changes on the cryosphere on multiple time scales. The need for such an dataset is well justified and the authors cite ample relevant literature. The paper is basically well-written and presented.

First suggestion: it is best to delete “Using the Google Earth Engine (GEE) platform” in the first sentence because GEE is just the platform of producing the product, it is not the main contribution of the study. However, I suggest to add a subsection in the section of “data and preprocessing” to describe the computing platform and the reason of choosing GEE.

Second question: why not add the dataset of 2020 year?

Third question: is it sufficient enough to validate so big product using just eight Landsat-5 images?

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

1. Line 17, NIEER is the abbreviation of the authors institute, “the new NIEER product” is difficult to project to the new AVHRR snow cover extent product, because many products are produced by the institute.
2. Line 22, “the producer’s accuracy was 81.0% the user’s accuracy was 81.3%, and the Cohen’s kappa value was 0.717” --> “the producer’s accuracy is 81.0%, the user’s accuracy is 81.3%, and the Cohen’s Kappa value is 0.717”.
3. Line 27, “nearly 40%” should be changed to precise number.
4. Line 37, “these decades” refers to which decades? The citing references was published in 2005 and 2018.
5. Line 38, “is not only of particular importance for climate research but also an indispensable indicator of climate change” needs to be rephrased.
6. Line 47, “throughout China” should be deleted, because they are global products.
7. Line 52, “The polar orbit meteorological satellite of” should be deleted.
8. Line 56, the abbreviation SCE has been described in the last paragraph.
9. Line 84, “processing” -> “preprocessing”.
10. Line 99, the sensor attenuation refers to AVHRR? Please specify it. Then, why the AVHRR sensor attenuation requires the algorithm to choose different TM images? And how different?
11. Line 101, “true values” -> “ground truth”.
12. Line 103, where is the main seasonally snow-covered areas in China? Please specify it. This means move the content at Line 122 here.
13. Line 124, the usage of the snow depth data for the proposed study (line 127) should be used as the first sentence.
14. Line 128, the cloud/snow confusion is generated by averaging the hourly ERA 5 land climate reanalysis dataset?
15. Line 137, “between” should be deleted.
16. Line 139, “forth” -> “fourth”.
17. Line 157, did you use the NDVI? If not, the formula (1) is not needed.
18. Line 175, “previous” -> “the previous”. It is the same in line 272.
19. Line 219, “algorithm above” -> “above-mentioned algorithm”, and “first” should be deleted.
20. Line 230, I suggest to add Huang as one of the co-authors.
21. Line 267-268, how to determine these thresholds?
22. Line 270, “Methodology” -> “Metrics”.
23. Line 291, what did cause the poor quality of raw satellite?
24. Line 315, delete “new”.
25. Line 319, delete “the”.
26. Line 321, “edges” -> “boundaries”.
27. Line 327 and 328, “snow-covered areas” -> “snow cover”.
28. Line 353, “poorly” is not good to describe the others’ product.
29. Line 375, what is the meaning of “GF”? Why not delete it?
30. Line 379 and 380, why not list the names according to the increase of the values?
31. Line 383, as an ESSD paper, the conclusion should focus on the data, not the method.

This paper employed a multi-level decision tree algorithm to detect cloud and snow based on AVHRR SR V4 data, then combined the HMRF-based spatio-temporal modeling technique and the snow-depth interpolation method to fill data gaps gradually. It produced a daily NIEER AVHRR SCE product with a spatial resolution of 5 km over China from 1981 to 2019. This product was validated using in situ snow depth measurements and SCE maps derived from Landsat-5 TM.

Since the cloud and snow confusion, as well as data gaps caused by cloud are common and occur often when mapping daily snow cover extent from optical sensors, techniques to solve these problems are valuable. The presented processing scheme is able to improve the quality of snow cover detected from AVHRR data. The produced long time-series snow cover extent product could be a significant dataset for studying climate change over China.

Despite of its significance, several issues still need to be resolved before a publication to ESSD. The quality control of AVHRR, why Landsat maps could be the true values to validate the cloud samples from AVHRR maps, and the three levels decision tree in flowchart could be sufficiently explained. In addition, it is not necessary to define a weight in HMRF modeling, since only one energy source was used in this study. Besides, the English of this paper should be further refined so as to improve the overall presentation.

Other comments and suggestions:

1. P1, Abstract, change “15 percent” to “15%”.

2. P2-3, it is suggested to provide a summery table of the mentioned SCE products, which lists the begin/end time, spatial resolution, temporal resolution, institution, produced methods, referenced paper, and download link.

3. Figure 1, it only shows 7 Landsat senses for validation. Change “Landsat Snow Maps” to “Landsat senses for validation”.

4. P6, it is not clear that how to deal with these night/ dense dark vegetation /sunglint/ water/ cloud shadow/ cloudy/ unused pixels, according to the quality control information.

5. P6, this paper used Landsat maps as the true values to validate the cloud and snow samples from AVHRR maps at the same days. It is okay for snow samples. However, for cloud samples, it depends on the overpass time of two satellites, since cloud can change in a quite short time period. Please provide more explanation.

6. Figure 2, it is not clear for the three levels decision tree. For example, how about the hierarchical relationships among them? How about the input and output for each level?

7. Figure 3, it is suggested to change the text color of “snow” to green, and that of “cloud” to orange.

8. Figure 4, it is suggested to change the text color of “snow” to green, and that of “Non-Snow” to brown.

9. P9, the original HMRF snow framework integrates spectral information, spatio-temporal information, and environmental information to reclassify snow and non-snow classes. The total energy function includes each energy source and its optimal parameters to minimize the total energy function. Among them, the parameter indicate the contribution of corresponding energy source. The original HMRF modeling technique employs a cubic spatio-temporal neighborhood to represent the combination influence from temporal context and the spatial context, which is effective to fill the overwhelming majority of data gaps in MODIS snow cover products. This research only used the spatio-temporal information, it is not necessary to define a weight for one energy source, as shown in equation 3. It is suggested to replace it by the spatio-temporal cubic energy function.

10. P10-14, it is suggested to add some more thorough analysis and discussion. For example, the accuracy over North Xinjiang, Qinghai-Tibet Plateau, and Northeast China, the accuracy over different land cover types, as well as analyses and discussion with previous studies.

11. Table 2, change “Year” to “Time period”.

12. Table 4, change “DEM<300” to “DEM≤300”. Please provide more detailed information about the cloud detection and the corresponding thresholds. What are the clues to divide the Target A into A1-A4 and to divide Target B into B1-B10? All threshold were determined by tests?

13. Table 5, please provide more information about the threshold values. How were they determined?

14. Table 10, the snow column of JASMES SCE should be near the Non-snow column of JASMES SCE.