General Comments:

This manuscript examines the impact of assimilating observations from various Radar Wind Profiler (RWP) networks on NWP forecasts of heavy rainfall in the Beijing–Tianjin–Hebei (BTH) region of China. These observation impact experiments are performed using an identical-twin observing system simulation experiment (OSSE) with only RWPs being assimilated. Results indicate that adding more RWP sites improves forecasts of heavy rainfall, with the configuration adding RWPs to both the ridge and foothills of the Taihang Mountains resulting in the largest benefits. Sensitivity tests indicate that the RWP impact is noticeably reduced when the detection altitude is reduced and when the vertical resolution is reduced. Interestingly, increasing the detection altitude past 8 km does not greatly improve forecasts. This information can be used in the future to help inform the further development of the RWP network in China.

Overall, the study provides a useful step forward in better understanding observation impacts on mesoscale precipitation events in China, and I thank the authors for putting in the time to perform this study and write up the results. I do have a few of general suggestions, though, that I think will help improve the manuscript:

1. It seems like the OSSE framework used here is an identical twin (i.e., the forecast system and truth simulation use the same model) and omits observation types other than RWPs. While this is not inherently a problem, it would be very helpful if the authors discussed the limitations of such a setup and used these limitations to qualify their results. For example, not including other conventional observations (e.g., radiosondes, surface stations) might inflate the impact of RWPs. Additional information about the limitation of such a setup can be found in Hoffman and Atlas (2016).
2. Some of the claims in section 4 do not appear to be fully supported by the figures. Can the authors revisit this section and double-check the claims made here? Some of the claims are also rather vague (e.g., on lines 433–436, it is not clear which metrics are being examined when claiming that one run is better than another run). Here are some specific examples of claims not fully supported by the figures (this list is non-exhaustive):
1. Lines 319–321: It looks like the RD and FH_RD experiments are also worse than CTL in the 550–500 hPa layer for u and that the RD and FH experiments are worse than CTL in the 700–550 hPa layer for v. So maybe it would be better to simply say that “DA experiments assimilating ridge and foothill RWPs generally outperform CTL”?
2. Line 499: There are some points on Fig. 15 that fall outside of the bias range listed here. E.g., 1-h, 30-dBZ CREF forecast for 20230712; 6-h, 40-dBZ CREF forecast for 20230721; 1-hr, 40-dBZ CREF forecast for 20230712.
3. Line 525: Based on Fig. 16b, it looks like all DA experiments actually increase the bias to make it farther from unity, not closer to unity.
4. Line 556–558: The text states that POD initially increases, then quickly declines, but based on Fig. 17a, POD appears to increase during the entire 6-h forecast.
3. The authors include some discussion as to why more RWP observations improve forecasts of heavy rainfall, which I think is really helpful. I think these discussions can be expanded, however, and made more clear. Some specific examples:
1. Line 32: The phrase “enhances mesoscale dynamics” is vague.
2. Lines 401–406: It is not clear what exactly is meant by “favorable environment for heavy rainfall”. Is this the result of increased convergence?
3. Lines 553–556: It is not immediately clear to me why (b) results in RWP observations from the foothills having a more immediate impact on forecasts than RWP observations from the ridgeline.
4. Line 578: Convergence zones are mentioned here, but nowhere else in the manuscript. If properly forecasting convergence zones is important to capturing heavy rainfall events in this area, it should be discussed in more detail somewhere in the manuscript.

In addition to these general comments, I also have several specific comments and technical corrections that are smaller in scope, which are listed below.

Specific Comments:

1. Line 42: Can the authors provide a range of vertical and temporal resolutions for RWPs here? I know this is included later, but I think it would be worthwhile to include it here to give the reader a sense of what is met by “high vertical and temporal resolution”.
2. Lines 64–81: There is a lot of discussion here about the geography of the Beijing–Tianjin–Hebei region. I’d recommend including some references to the map in Fig. 3 so that the readers have a better sense of where exactly the mountains are relative to Beijing and what the density of the operational RWP network looks like.
1. Line 124: A reference to Fig. 3 would also be helpful here. Does the domain of Fig. 3 perfectly match the WRF domain? If not, can the WRF domain be added to Fig. 3?
3. Lines 96–99: Can a citation or two be added here to back up this claim? 
4. Lines 156–158: Is there any DA in the truth run? My guess is that there is not (as is typically the case in OSSEs), but including the phrase “DA cycling” in this sentence makes it sound like the truth run might include DA. Can the sentence be reworded so that it is clear that the truth run does not include DA?
5. Fig. 2: It is difficult to see the rain gauge observations in this figure, and it is also difficult (at least for me) to differentiate between the different dot sizes. Can the rain gauge data be plotted in a different manner to make it more clear? One idea: perhaps another row or column can be added to this figure that shows rain gauge observations as color-filled dots?
6. Lines 205–207: What exactly is the NOAA topographic dataset being used for here? Is this how RWP site elevations are determined?
7. Fig. 3: Can the Bohai Bay be made a different color to make it clear that this is a large body of water and not just flat land? Similar to terrain differences, differences in land cover between land and water can have a large impact on local meteorology.
8. Lines 224–225: There are certainly several similarities between the forecast system used here and WoFS (e.g., both use a very similar dynamical model), but the DA in WoFS seems very different. As mentioned in the manuscript, WoFS relies heavily on flow-dependent background error covariances derived from an ensemble (either in an EnKF or hybrid 3DEnVar), which is a notable difference from the pure 3DVAR approach used here.
9. Lines 227–228: Can the authors confirm that the majority of RWP observations were collected in regions without precipitation? If not, does removing the RWP observations in regions of precipitation change the results?
10. Lines 229–230: Which GFS forecast hours are used for the initial and boundary conditions?
11. Lines 255–263: Based on this section, it sounds like there are no cross-variable correlations between RWP winds and other state variables (i.e., assimilating RWPs only updates the winds and not other variables like temperature and humidity). Can the authors confirm this? How will this impact the results from this study?
12. Line 256: Do the background wind errors increase or decrease with height? It is not clear from this sentence.
13. Line 286: Given that the model grid spacing is > 1 km, the term “convection-allowing” is more appropriate than “convection-resolving” because the convection is not fully resolved (e.g., Potvin and Flora 2015).
14. Lines 300–302: I am having trouble following the second part of this sentence. Maybe it can be reworded?
15. Line 377: This sentence is a bit confusing. Is the message that the intensity errors are reduced in the FH experiment, but there is still a positive bias?
16. Lines 398–399: By “initial,” are the authors referring to the 1-h APCP forecasts?
17. Lines 492–512: This section only compares the FH_RD experiments between the cases, but it seems like it might be more beneficial to compare each of these FH_RD experiments to their respective control runs. This comparison would provide a much clearer picture of how RWP observations improve NWP forecasts.
18. Lines 525–526: How is statistical significance determined here? Based on Fig. 16, statistical significance can not be deduced because there appears to be overlap between the confidence intervals at all times. This does not necessarily mean that the difference in forecast statistics is not statistically significant (e.g., see 5.4.6 in this reference from Vanderbilt University: https://researchguides.library.vanderbilt.edu/c.php?g=69346&p=855555#), but we can not say for certain that the difference is significant without another type of statistical test. If another test has not been performed, I would recommend removing this mention of statistical significance.
19. Code and data availability: Can the authors also provide a link to the DA code used in this study? It would also be really helpful if the authors could provide the namelist files for WRF and the DA in a public-facing repository.

Technical Corrections:

1. Line 20: I’d recommend replacing “in rainfall forecasting” with “for rainfall forecasting”.
2. Line 31: “benifiting” should be “benefiting”.
3. Line 41: “state-of-art” should be “state-of-the-art”.
4. Lines 55–57: This sentence sounds awkward, particularly the phrase “over 100 sites are deployed by 2020”. Can it be reworded?
5. Line 59 (and elsewhere): Definite and indefinite articles (i.e., “the” and “a”) are missing in several locations in sections 1–3. More information about when to use definite and indefinite articles can be found here: https://owl.purdue.edu/owl/general_writing/grammar/using_articles.html. Here are some specific examples (italics indicate an added article; this list may not be exhaustive):
1. Line 59: “above national average” should be “above the national average”.
2. Line 73: “in BTH region” should be “in the BTH region”.
3. Line 76: “like Taihang mountains” should be “like the Taihang Mountains”.
4. Line 80: “of RWP network” should be “of the RWP network”.
5. Line 80: “in Taihang Mountains” should be “in the Taihang Mountains”.
6. Line 99: “prior study” should be "a prior study”.
7. Line 106: “over operational RWP network” should be “over the operational RWP network”.
8. Line 114: “of NWP model” should be “of the NWP model”.
9. Line 138: “enhance wind field analysis” should be “enhance the wind field analysis”.
10. Line 139: “in PBL” should be “in the PBL”.
11. Line 220: “distribution of operational RWP network” should be “distribution of the operational RWP network”.
12. Line 221: “ridge of Taihang Mountains” should be “ridge of the Taihang Mountains”.
13. Line 231: “hour of analysis” should be “hour of the analysis”.
14. Line 237: “end of DA cycles” should be “end of the DA cycles”.
15. Line 274: “using neighborhood radius” should be “using a neighborhood radius”.
6. Line 64: “Heibei” should be “Hebei”.
7. Line 69: “plain” should be “plains”.
8. Line 72: Is there a reason why “large” is used instead of “synoptic” here?
9. Lines 75–76: Add the word “in” between “network” and “concentrated”.
10. Line 78: “observation” should be “observations”.
11. Line 81: “forecast” should be “forecasts”.
12. Lines 89–92: Can this sentence be reworded? In particular, the phrases “few peer-reviewed published research” and “RWP network associated with complex terrain” (which also appears on line 93) are confusing. Perhaps the latter could be rewritten as “a RWP network in complex terrain”?
13. Line 106: “Does” should be “Do”. 
14. Line 122: Remove the word “space” after “1.5-km”.
15. Line 136: Remove “and” and replace with a comma.
16. Line 136: “atmospheric motion vector” should be plural.
17. Line 141: Remove comma after “3DVAR”.
18. Line 143: Add “enhancements” after “RWP network”.
19. Line 168: Remove “of” between “southwest” and “Beijing”.
20. Lines 262–263: Please rewrite so that this sentence does not start with “And”.
21. Line 269: “experiments” should be singular.
22. Line 298: “uncertiaties” should be “uncertainties”.
23. Line 314: It looks like the analyses are solid, not dashed.
24. Fig. 6: Note that the noDA experiment is black in the caption.
25. Line 335: Add this sentence to the caption of Fig. 7.
26. Line 410: A different word should be used in place of “estimates”. The rainfall from the truth run is the actual rainfall amount for the OSSE, not an estimate.
27. Line 413: "accumulate" should be “accumulated”.
28. Line 501: The transition “however” sounds out of place here.
29. Line 509: The transition “nevertheless” sounds out of place here.
30. Line 664: “accsible” should be “accessible”.

References

Hoffman, R. N., and R. Atlas, 2016: Future Observing System Simulation Experiments. Bull. Amer. Meteor. Soc., 97, 1601–1616, https://doi.org/10.1175/BAMS-D-15-00200.1. 

Potvin, C. K., and M. L. Flora, 2015: Sensitivity of Idealized Supercell Simulations to Horizontal Grid Spacing: Implications for Warn-on-Forecast. Mon. Wea. Rev., 143, 2998–3024, https://doi.org/10.1175/MWR-D-14-00416.1. 

The study investigates the benefits of using radar wind profiler (RWP) observations for forecasting convective initiation (CI) in small-scale boundary layer convergence zones.  The research employed the Weather Research and Forecasting (WRF) model along with the NSSL3DVAR data assimilation (DA) system.  Synthetic RWP data, generated through Observing System Simulation Experiments (OSSE), were assimilated for three summer heavy rainfall events in the Beijing-Tianjin-Hebei region. The results indicated that assimilating RWP observations improved model initial conditions and enhanced short-term severe weather forecasts.  Notably, improved forecasting outcomes were observed when combining operational, foothill, and ridge RWP data.  Multiple sensitivity experiments were conducted to evaluate the impact of vertical resolution and maximum detection heights.  The study identified optimal configurations for future real-time RWP data assimilation.

The research and results of the study are both interesting and significant. The manuscript is well-structured and clearly presented, offering a thorough analysis using the OSSE method. It is well-written and will be ready for publication after a few minor revisions and responses to my questions.

Comments:

1. In lines 144-145, ‘so we did not use the ensemble derived background error covariance, which is also incorporated in the variational framework’. What method is used to compute the background error statistics? What control variables are utilized to calculate the B-matrix? Is the covariance matrix a day or wet matrix. Write a few details about the calculation of the B-matrix. 
2. Line 135, ‘total precipitable water' are the products or sources of observations used for assimilating total precipitable water? The author needs to include appropriate references for each observation.
3. Authors need to include additional information about the NSSL3DVAR assimilation system. This should encompass the minimization cost function, control variables, and the calculation of the B-matrix.
4. Line 132: Add a few references after the phrase water path, total precipitable water, and atmospheric motion vector.Jones, T. A., Wang, X., Skinner, P., Johnson, A., & Wang, Y. (2018). Assimilation of GOES-13 imager clear-sky water vapor (6.5 μm) radiances into a warn-on-forecast system. Monthly Weather Review, 146(4), 1077–1107. https://doi.org/10.1175/MWR-D-17-0280.1        Mallick, S., & Jones, T. A. (2020). Assimilation of GOES-16 satellite derived winds into the warn-on-forecast system. Atmospheric Research, 245, 105131. https://doi.org/10.1016/j.atmosres.2020.105131
5. Did the authors use any additional conventional observations or satellite observations in their assimilation system?
6. Line 201, rewrite the line ‘meridional wind components (u and v)’ to ‘meridional wind (u and v) components’.
7. Line 217, ‘The perturbations are assumed to be normally distributed Gaussian random errors with a mean of zero and a standard deviation of 2 m/s’ , How do you calculate the observation error for each location?
8. In the summary section, write a few words on  how incorporating flow-dependent background error covariances in data assimilation (DA) systems can enhance precipitation forecasting compared to the static background errors used in 3DVAR.
9. Based on OSSE results, what are the key considerations for designing and deploying optimal RWP networks in complex terrain regions?