This paper decribes the effort to develope a database based on historical radiosonde intercomparison activities. Knowing the instrument and correction information is the key to establishing a reliable upper air time series for long term climate change study.  However, information for historical radiondes is often fragment and incomplete.  I understand that collecting the data and compiling a database is a tedious task.  This is an excellent work.  It can be very helpful for the global radiosonde archive, thus for the global change study. I'm looking forwarding to seeing the database to be published in its final version.

The paper is well organized; data backgroud and database structure have been clearly illustrated.  There are a lot more information in the historical upper air observations but I feel the important and necessary information is presented by the authors.  I only have some minor points here:

1) L39-40: data base--> database

2) L42: use of it-->use of them

3) L57-58: the transmission of the data and their processing--> data transmission and processing

4) L66-67:  Were the soundings on International Days and Week launched at the same UTC hours?

5) L74-75: Payerne (Switzerland) --> Payerne, Switzerland

6) L80: Shlyakhov. -->remove the dot at the end

7) L92-93: "GCOS (Global Climate Observing System), the GRUAN (GCOS Reference upper-air network) " maybe changed to: "Global Climate Observing System (GCOS), the GCOS Reference Upper-Air Network (GRUAN)"

8) L93-94: and with this the quality is --> with which the radiosonde quality was

9) L102: amount of studies on radiosondes has been --> number of studies on radiosonde have been

10) L117: "which is useful to determine, were corrections have been applied." , this sentence is not clear.

11) L130: the 1930 and 1990 --> change to "the 1930s and 1990s" or "1930 and 1990"

12) L148: Figure 1: Left: Three types of Väisälä radiosondes --> Figure 1: Three types of Väisälä radiosondes.  Left:

13) L226: have been-->were

14) L227: deviate--> differ

15) L230: "resp.", do you want to say "or"?

16) L266-267: "the individual campaign" --> individual campaigns

17) L267: unique identifier (UID)

18) L283: relate-->related

19) L285: have-->has

20) L291: end of line ":"--> "."

21) L335: intercomparison-->intercomparisons

22) L340: report-->were reported

1. L93 on GRUAN can include “Bodeker et al. (2016). (see the list of GRUAN publications on https://www.gruan.org/documentation/articles).
2. L95-L100: Dirksen et al. (2014 for RS92) includes all those methods and should be mentioned.
3. L11 & L23: “climate” should be added here too.
4. L29: Those two references should be mentioned.

Zhou, C., J. Wang, A. Dai and P. Thorne, 2021: A New Approach to Homogenize Global Twice-daily Radiosonde Temperature Data from 1958 to 2018. J. Climate, pp.1-64.

Dai, A., J. Wang, P. W. Thorne, D. E. Parker, L. Haimberger, and X. L. Wang, 2011: A new approach to homogenize daily radiosonde humidity data J. Climate, 24, 965-991.

5. In term of how reliable the error assessment is, it would be good to explain some of outliers in Figure 6, such as two in Fig. 6d.

6. This is a really nice archive of results from early radiosonde intercomparison campaigns. However, I think that the main application of the study is to help adjust errors in early radiosonde data, so the data can be part of homogenized radiosonde data for climate studies, esp trend analysis. A lot of prior work including those from the second author have devoted to homogenize global radiosonde temperature data. It would be good to provide some examples on whether the adjustments made from prior homogenization efforts are consistent with the errors archived by this work. If not, how does it affect the conclusions made by those studies? This would greatly highlight the usefulness of this archive.

7. The examples given in the manuscript are all temperature data. Is it also true in your database? If so, it should be clarified.