The manuscript entitled “Generation of global 1 km all-weather instantaneous and daily mean land surface temperature from MODIS data” has been reviewed. The authors generated a global all-weather instantaneous and daily mean LST product spanning from 2000 to 2020 using XGBOOST, and then they systematically evaluated the accuracy of the produced LST dataset. The manuscript is written well and can be easily understood. I think the manuscript can be accepted if the following concerns have been answered well.

1. Several most recent studies corresponding to all-weather LST datasets have been published on the ESSD, but they are not cited in this manuscript. Although the current study focuses on a global scale. Some regional/national-scale studies on all-weather LST datasets also have important contributions and need to be mentioned in this manuscript.
2. I am concerned about the effects of station density on the model accuracy. Please include some analysis on this concern.
3. I am especially concerned about the spatial pattern of the estimated LST data. The cloudy-sky LST may be mainly decided by the ERA5 data (coarse spatial resolution), which may lose some spatial details. Figs. 13 and 14 have shown the spatial pattern of LSTs. However, the terrains on the two selected tiles may be not representative. 1) Does the estimated LST can show spatial details in mountainous regions? 2) Meanwhile, can the urban heat island effects be shown clearly? 3) Are the estimated LST data spatially and naturally smoothing without any abnormal boundaries?
4. I would like to check more daily mean LST data, instead of the daily mean LST on the first day of a year, nor the monthly-scale LSTs. The shared LSTs are not representative of quality checking by reviewers.
5. Why not include LST data in 2021 and 2022?

This study developed a global 1 km all-weather instantaneous and daily mean land surface temperature dataset. The structure of the manuscript is clear, and the contents are abundant. I think this manuscript can be published after major revisions.

1. Introduction. Authors should cite more recent literature from the past 3 or 5 years.
2. Lines 40-44. The authors mentioned that both satellite and station data are utilized in various fields, but only cited papers based on satellite data.
3. Lines 51-60. The orbit gaps in MODIS data also result in data gaps.
4. Why not utilize land cover and vegetation index? Both variables are strongly correlated with land surface temperature.
5. Section 2. The obtainable years for all data and the years of data used in this study were not mentioned.
6. The texts in some figures (e.g., Figure 1) are too small to be seen clearly.
7. Table 1. Please provide the access link for obtaining this data. It is suggested to include the data from Section 2.2 and 2.3 into Table 1.
8. Line 325. Resampling low-resolution data directly to high-resolution may affect the estimation of land surface temperature.
9. Lines 323-334. Why are these two paragraphs nearly identical?
10. Line 352. Why does the data cover the time range from 2002 to 2018? The time range in abstract is from 2000 to 2020.
11. Section 3.3. Could you please better clarify the novelty of and innovation of your method?
12. Figures 4 and 6. Unit should be added. Figures 4-6. The legend should include the data.
13. Lines 421. This explanation is far-fetched. The reason might be the greater spatial and temporal differences in daytime land surface temperature compared to nighttime land surface temperature.
14. Lines 420 and 447. References should be added to support the explanation.
15. Section 4.2.3. Why not compared with other similar products developed in previous studies?