The manuscript entitled “Investigation of near-global daytime boundary layer height using high-resolution radiosondes: First results and comparison with ERA-5, MERRA-2, JRA-55, and NCEP-2 reanalyses” presents a near-global assessment of high-resolution radiosonde derived boundary layer height (BLH) and provides a quantitative assessment of four reanalysis products. This paper is generally well written and makes an important contribution to characterizing the BLH at the global scale and providing useful information on reanalysis data usage. However, I have the following major comments concerning the bias attribution.

First, in the case study at Chongqing, the fine-scale vertical structures of Ri, WS, RH, and T seem to have a larger impact in determining BLH compared to the overall bias of the basic parameters. It appears that both overestimation (in JRA-55) and underestimation (in NCEP-2) of WS and RH lead to a smaller BLH. Discussions on the impact of vertical structure including the vertical resolution would provide useful information on the bias attribution.

Relatedly, is there a specific reason for choosing this case as an example to show BLH biases in the reanalysis data? It would be helpful to provide a comment on other cases.

Second, the biases of the BLH in reanalysis data are attributed to the complex topography and static stability based on their correlation coefficient. The afternoon sounding during the warm season leads to large biases over the TP and western US, where the terrain is complex. Assessing the relationship between BLH bias and DEM spread using data collected at similar LST would provide useful information on the robustness of the results.

Meanwhile, because of the coarser temporal resolution, MERRA-2, JRA-55, and NCEP-2 are not able to match LST of all soundings during IOP. The time mismatch between the sounding and reanalysis data may also introduce biases due to the distinct diurnal variation of BLH. It is necessary to discuss if the result will significantly change with/without IOP data used.

Fig. 4 nicely shows the diurnal variation of BLH. The authors mention “some soundings that are released at 0000 and 1200 UTC are excluded …. for collecting samples in the daytime.” In my understanding, for instance, the 14 LST results in both Fig.4a and Fig. 4b should include all soundings collected at 14 LST. It is not very clear why some soundings at 0000 and 1200 UTC are removed to only show daytime results in Fig. 4b?

Besides, how does the application of additional soundings during IOP lead to the differences between Figs. 4a and 4b?

Is there a specific reason for presenting the difference using radiosonde (the reference dataset) minus reanalysis rather than reanalysis minus radiosonde in Figs 5-8? It seems counterintuitive to use positive differences in those figures to represent underestimated BLHs.

Specific Comments:

Line 56: Suggest changing to “boundary layer height”.

Line 192: How many layers below 500 hPa in ERA-5?

Line 218: Change to 0000 and 1200 UTC.

Line 220: This section introduces calculations for both normalized sensible heat and latent heat fluxes. Suggest changing the section title to include both fluxes.

Line 225: Add a period after the parenthesis. Can you further explain why small latent heat flux means more energy being available for PBL growth?

Line 236: Remove “sensible”.

Line 237: Sections 2.4 and 2.5 introduce BLH calculation which may be more connected to section 2.2. Suggest moving those two sections forward.

Line 272-273: Is this an extra step only required by observations during IOP, as the regular synoptic times are included in all reanalysis data? Meanwhile, JRA-55 and NCEP-2 have a temporal resolution of 6 hours, which may be not able to hit every weather balloon launch time with hour difference. Would it result in a significantly smaller sample size compared to ERA-5 and MERRA2?

Line 282: Is there a specific reason for arranging the panels in the order of a, b, d, c?

Line 345-346: The authors mentioned that the reanalyses and observations show the deepest BLH in the afternoon of summer, from which I think it is insufficient to conclude that “both capture the diurnal and seasonal variations” at this point.

Line 365-366: Did the authors mean “latitude” and “67.6 °N/°S”?

Line 385: Remove “/.”.

Line 392: What is the “ensemble mean”?

Line 413: Change “WD” to “WS”, and at other places.

Line 423: Fig. 9b marks significant correlations between BLH and Ps. I think this was simply left out by mistake.  

"Investigation of near-global daytime boundary layer height using high-resolution radiosondes: First results and comparison with ERA-5, MERRA-2, JRA-55, and NCEP-2 reanalyses" provides validations of simulated boundary layer heights from four commonly-used reanalysis products on a near-global scale. The manuscript is nicely organized and comprehensive. Given the important role of reanalysis products in climatological analyses, in energy-focused resource assessments, and as inputs to higher-resolution models, validations such as the one presented here are essential for understanding reanalysis biases and limitations. Comments and suggestions for enhancement of this manuscript follow.

The discussion on the vertical resolution limitations of IGRA and the reanalysis products (Lines 96-99) would improve by including the numerical vertical resolutions (exact, on average, or a range) for each of these products, instead of simply stating that they are sparse. This information is provided in Section 2, but since the manuscript defines the resolution of GPS RO on Line 92 it would be helpful for comparison to have this information for IGRA and the reanalyses in this location as well.

The authors are disregarding the packaged BLH parameter from MERRA-2 and recalculating BLH in a more similar fashion to the ERA5 definition (Lines 202-207). For the benefit of reanalysis users, it is highly recommended that the MERRA-2 packaged BLH parameter is also validated along with the author-derived version. Can this comparative analysis be included?

Figure 1 is a helpful case study to assist the reader in the methodology. Could ERA5 be included in the graphic as well, instead of a brief mention in the figure caption?

Section 3.3 provides an interesting attempt to correlate BLH with near surface measurements, however it should be moved to a different location in the manuscript, as it does not involve the reanalysis products and therefore does not flow with the surrounding sections. Additionally, the enthusiasm over the correlation results in this section should be tempered. 0.39 is not a "relatively high positive correlation coefficient". Perhaps "moderate" might be a better choice.

I second Anonymous Reviewer #3 in suggesting that presenting results according to reanalysis minus radiosondes is much more easily understood that radiosondes minus reanalysis.

Specific comments:

Line 33: Suggest adding "analysis" after "air quality, weather and climate".

Line 85: Suggest rewording "And notable diurnal and seasonal cycles have been revealed" to Notable diurnal and seasonal cycles in BLH variation have been revealed".

Line 113: Elaborate numerically on "a rough consistency".

Line 184: Suggest rewording "As a result, ..." to "Using this definition, 190,013 profiles including soundings launched at both synoptic times and during IOP, spanning January 2012 to December 2019, are used to obtain the BLH in the daytime."

Line 190: Reword "undergo" to "has undergone".

Lines 205, 382, 384, 460: Change "MERR-2" to "MERRA-2".

Line 213: Give NCEP-2 its own paragraph beginning here, as was done for the other reanalysis products.

Line 363 and Figures 5-8: "Austria" should be "Australia"?

Line 377: Recommend numerically describing the seasonal differences in the bias.

Line 394: Reword "acceptable" to "more in line with the observations". Acceptable is too subjective.

Line 425, Figure 11: Change "EAR-5" to ERA5.