The maper "A micro-genetic algorithm for combinatorial optimization of physics parameterizations  in  Weather Research  and  Forecasting  model  for quantitative precipitation forecast in Korea", by S. Park and S. K. Park, analyzes the importance of optimizing the selection of the parameterization physics schemes, and their parameters, in improving the prediction of extreme rainfall at the mesoscale in Korea.

The paper is generally well written and carefully describes the performed experiments.  The paper appears scientifically sound and the degree of novelty is high. However there are some small points to be improved before publication, in my opinion.

i) The paper is based on the comparison between several simulations performed with oprimized WRF and a single meteorological event. However, the event is only shorthly descrived (section 4.1). There are two reference: KMA (2018) shows a page in Korean in which, for a foreigner, is not easy to understand (and in any case does not contain a meteorological description of the event). Park and Park (2020) is another paper. In my opinion, a paper should be self-consistent and contain all indication to undestand the problem; reference to other papers could be used only for details. For instance, Sokcho was the location showing the largest precipitation? During the 27 hours of rainfall there, there were some time periods in which rainfall was more intense?

ii) Maps in figures 5 and 8 show the precipitation in the second domain for all simulations. Since the portion of Korean territory (and adiacent sea) interested by the event is much smaller than the domain, I suggest to zoom on the portion of territory interested by the rainfall (about a quarter of the domain), in order to highlight the details. The other part of the territory is not important in this sense, since there was no any precipitation. In this way, it could be possible to better appreciate the structure of the precipitation area.

iii) it is very clear that this result appears quite interesting, since it shows the importance of using an accurate choice of the physics parameterizations schemes. However, this result could be dependent on the case study. In this sense, if possible, it could be interesting for the reader to add, in the discussions, a sentence in which there is an attempt to understand why some schemes perform better than others, from a physical point of view.

Language: I am not a native speaker, so my opinion on the language, very good, could be biased. There are some minor typos to be corrected (e.g .: gird -> grid, Ninno -> Niño).

This study used the micro-genetic algorithm and performed a combinatorial optimization of cumulus, microphysics, and planetary boundary layer schemes from the WRF model to improve the quantitative precipitation forecast of a heavy rainfall event in Korea. Five different fitness functions are chosen in terms of Equitable Threat Score to obtain the corresponding best combination of physics schemes for each fitness function. The methodology adopted in this study is explained in great detail, and this method could be very useful to obtain the best combination of physics parameterization schemes with a minimal number of model runs instead of performing an exhaustive search. The manuscript is well written in general, and the results are presented with clarity. 

Having said this, I have a few comments on the manuscript.

1. The main concern with this work is that the results obtained from this study are very specific to the heavy rainfall event that was considered. Recent studies on parameter estimation, such as Duan et al. (2017) and Di et al. (2018), have performed simulations over multiple heavy rainfall events to obtain a set of optimal parameters to improve the precipitation prediction. Since only a single event was considered in this study, the robustness of the results obtained could not be ascertained. It would be good if the authors could repeat this exercise for a few more heavy rainfall events and present a comprehensive result. The authors should at least try to check how the optimum schemes (OPT-EXP1) are performing for other heavy rainfall events compared to REF and CTL. 
   
   

   
   
   Duan, Q., Z. Di, J. Quan, C. Wang, W. Gong, Y. Gan, A. Ye, C. Miao, S. Miao, X. Liang, et al. (2017). Automatic model calibration: A new way to improve numerical weather forecasting. Bulletin of the American Meteorological Society, 98(5), 959–970.

   Di, Z., Q. Duan, C. Wang, A. Ye, C. Miao, and W. Gong (2018). Assessing the applicability of WRF optimal parameters under the different precipitation simulations in the Greater Beijing Area. Climate Dynamics, 50(5-6), 1927–1948.

2. In Figures 6 and 9, the Correlation coefficient, R, was mentioned in the scatter plots. But some values of R were more than 1. As the values of the correlation coefficient must lie between -1 and +1, the authors should check the calculation of the correlation coefficient.  
3. In Fig.1, The positions of Yes and No seem to have been swapped in the outer loop. As mentioned in Section 3.2, if the maximum number of generations is exceeded, the algorithm should stop.
4. Line 194, “...referred to as OTP.” I think the authors meant OPT.
5. The language of the manuscript needs to be improved. There are some small spelling and grammatical mistakes in some places. Some of them are: Line 196; “grid” is written as “gird”, Lines 235 and 345; “12th” is written as “12nd”, Line 236; “4th” is written as “4rd”, Line 238; “physics” is written as “physic”