< General comments >

This article reviews and verifies the importance of radiosonde observation data in higher altitude beyond 10hPa with several demonstrations including numerical experiment. In general, it is well written and the article concept is clear. Numerical experiments described in Section 5.2 demonstrated well for the importance of higher altitude observation needs.

However, I wonder why this paper does not explicitly mention or provide any necessary (or desired) quantities for climate research. I mean there is no description about necessary accuracy, observation frequency, geographical distribution and so on. Of course, I can easily guess that it depends on the targets, so it might be difficult to identify or answer. In addition, actually, as to the accuracy, the authors referred the WMO’s statement related to GBON (L. 97-109) and OSCAR requirements, and this article notes the recent radiosondes are satisfied with this. But, I wonder such accuracy should be considered in conjunction with temporal/spatial distribution in terms of climate research (or weather forecast).

In short, I’d like to know the goal that the authors aim by attaining data in higher-altitudes.

For example, as for the accuracy, the WMO’s statement and/or OSCAR are the basis?  The current GRUAN stations can provide enough number for the latter two (frequency, location)? Since it is expected that this article can act as a basis for future higher-altitude radiosonde soundings and this can solicit not only operational sites but also research sites to do such observations , information or any suggestions are helpful for their justification.

Below are specific comments, but I believe most of them are minor ones.

< Specific comments >

• L. 253-273; In order for accurate description, it is better to mention that there are two types of wind measurements using GNSS/GPS; one measures the wind using the Doppler shift of the GPS frequency, and the others calculate it from difference of locations between the measured intervals.
• L. 447-452; It might be useful to refer the recent hot topic “relationship between QBO and the Madden-Julian Oscillation,” as it demonstrates stratospheric phenomenon (QBO) might control tropospheric phenomenon (MJO), which has great impact for the global weather forecast.
• L. 585; Please spell out “RO” as it appears first time here.
• L. 634-635; After all, this accuracy is enough or not for any climate research?
• L. 690; “These show … below 10hPa.” should be inserted not here in this caption but in the text.
• Figs. 11 and 12; I wonder if only datasets, which contain higher (>30km) data, are used(*), mean and SD show similar tendency? Here, “similar” means for profiles above/below 30km. I thought It is helpful to know whether any effect from sampling exists or not. (*) … Namely, it means the calculation with the same number of pairs in the entire range (heights).

• L. 836-839; Isn’t it possible to express “high-” in quantitatively?
• L. 887-891; Could you explain a little bit more about “overall”? For example, if large RMSE is found in temperature for ECMWF (but others are small) and in wind for observations, they are evaluated on the same level?
• Figs. 15 and 17; Did the authors confirm any geographical bias for lower/higher soundings to avoid any possible their effect? (I mean, for example, Case A frequently does not appear in any latitudinal band?)  Also, any bias for sampling time zone (00Z or 12Z) for the Cases A and B?

• L. 1120; (Middle) should be (Right).

This is an important manuscript, highlighting the benefits of letting radiosondes fly higher into the stratosphere. It is well thought through and has an important message. It should be published, perhaps even featured in the journal. The text is high quality, strongly backed up with a large number of references. I found no typos or unclear formulations beyond those noted by the other reviewer. In my view the manuscript could be published in its present form

I have only one suggestion which would require some work if implemented:

The authors try to be as comprehensive as possible, and consequently it covers many aspects of the impact of radiosondes reaching high levels. The completeness also makes the paper a rather long read, though.  When trying to assess the impact on NWP the paper really leaves the field of measurement technology and inter-calibration and this section also has a different character. At this point the editor may suggest to split the paper into two, as the data assimilation aspects could as well be covered in a separate paper. On the other hand the operational NWP aspect may be important to convince meteorological services and funding agencies to raise the balloon burst height at more sites, not just at the GRUAN and GBON sites.

If it were decided to split the paper, one could perhaps include denial experiments where even more pressure levels are omitted from radiosonde profiles. In the 1950s and 1960s many Russian and Chinese radiosondes rarely reached 100 or even 200 hPa. This would certainly be interesting in the context of reanalysis applications.