This article proposes a parametrization for the diagnostic of a minimum value of eddy diffusivity. The goal is to improve the modelling of PM2.5 in the stable boundary layer. The scheme is applied to a test case in China with the WRF-chem regional meteorological model. They diagnosed that model generally underestimates weak turbulence in the nocturnal stable boundary layer leading to an overestimation of surface aerosol concentrations. To parameterize this Kzmin, they propose to use the sensible H and latent LE heat fluxes. The additional term increases the Kzmin value and enables to reduce significantly their model bias. They show that their change includes a spatial variability of the Kzmin, depending on the landuse and the meteorology. A latitudinal effect is diagnosed.

The fact to consider that all models overestimate surface concentrations of particles is not correct. It is the specific case of this model WRFchem. But there is no systematic tendency about this point. If the authors are sure of that, please provide the bibliography, a review article. But considering this is mostly a problem with this model, the fact to add a term to reduce this bias is a tuning. Except if the new term has a robust physical basis. For the moment, this additional Kzmin is able to unbias the model but it could be only an error compensation. Of course, it is alays difficult to quantify but the authors should at least discuss this point and add more sensitivity tests: injection of anthropogenic emissions at levels higher that the surface level, test of boundary layer scheme to see the model sensitivity to the bias of T2m, among others. Adetailed analysis based on hourly time-series and comparison to surface observation of PM1.5 is also missing to really see if there is a physically improvements of the surface concentrationsw with this scheme. Ideally, lidar data could help to see if the vertical structure of aerosols is better reproduced by the model. Another question: the comparison is only perfomed for PM1.5. What about PM10? NO2 and O3 (often measured at stations)? The bias on these species should be of interest to understand if the new Kzmin value is really better for all species.

Abstract:

The abstract is clear and summarizes correctly the whole content of the study, altough the last three sentences deserved to be reformulated.

2. Data and Metodology:

Several data are used to validate the model's hypotheses made in this study. It includes vertical measurements, essential for this type of vertical mixing study. The model used is WRFchem, known as a fully coupled model. Unfortunately, the coupling is not always really activated, all options being not coupled. It is recommended to the authors to add in Appendix an explanation about their namelist to ensure that the coupling was really fully active. The choice in the namelist may completely change their results.

The key point of the study is to assume that the 'evaporative fraction' (EF) may be used to characterize the searched value of Kzmin. Why not, but why exactly? 

The end of the paragraph (lines 130 to 137) is not very clear and should be reworded. It means that under stable conditions, the values of Kzmin may be 100 times larger than under unstable conditions?

3. Results and discussion

If some metrics are well known, please define them, including the IOA Index Of Agreement.

l.145: if the key point is an enhancement of Kzmin during the night, why not show time series of a few days, at a station with measurements and with hourly values, showing three to four consecutive days?

The Table  shows that the bias for T2m is -1.15 degrees when the text says -0.86 degrees. Please correct the correspondence between text and tables. And a bias of 1.15 degrees is not really good. There is perhaps a problem with the use of the WRF model, independently of the Kzmin parameterization studied in this paper.

l.165: the authors diagnosed a positive bias in PM2.5 surface concentrations and conclude it is due to "geographical conditions, climate and emissions differences and the degree of pollution". It is not an in-depth analysis. Before tuning one parameter, it should be useful to erform some sensivity tests in order to see if the bias is due to emissions, meteorology, transport, deposition of a mixing of all.

l.177: What new PBL scheme? The new Kzmin formulation? or something else? This sentence seems out of place in the text.

l.187: There is no quantified improvement but "we believe that the simulation in the YRD has also been improved." Please explain better.

Figure 4: Usually, measurements are with symbols and model outputs with lines. Profiles are very small and difficult to read.

The choice to have a maximum possible value of 2.0 is not a result but just an arbitrary threshold choice (l.135). Then on the map in Figure 5, some values may be larger than 2.0 (7% of the values).

l.220: "We need to use a larger Kzmin value to enhance turbulence under a strong stable atmosphere and small or no adjusted Kzmin values under a weak stable or neutral atmosphere."

Here there is an explanation of the choice made for the new Kzmin. But there is no physical proof of this choice.

l.223: Please use carefully the term 'climate' (not correct in the present context).

l.232: It is stated that: "As most primary pollutants are emitted to the first level in the model", it is not the case of many models. For example in Europe, anthroogenic emissions are vertically distributed following a vertical profile proposed by EMEP. Fires and dust emissions are often injected following a vertical profile.

In this study PM25 have different origins. Can a sensitivity experiment diagnose what source (anthropogenic? fires? biogenic? dust? etc.) could be responsible of the observed bias close to the surface?

“Parameterized minimum eddy diffusivity in WRF-Chem(v3.9.1.1) for improving PM2.5 simulation in the stable boundary layer over eastern China” by Lu et al. proposed a parameterization formula for minimum turbulent diffusivity (Kzmin) and tested the simulations effects for PM2.5. The results show that the revised Kzmin parameterization formula improved the PM2.5 simulation by improving turbulent diffusion under stable conditions. Weak turbulence in SBL is the key challenge in restricting progress of SBL theory and simulation, the topic of the manuscript is very important. However, the physical logic of the revised Kzmin parameterization formula is questionable, and numerical experiments need to be added. Therefore, I recommend major revision.

1. Line 45, the SBL, weak turbulence and turbulence intermittency are hot topics in studies of atmospheric boundary layer with a lot of papers and progresses, I suggest the citations here  keep up with the latest developments.
2. Line 49, “Huang et al. (2010)” did not show in Reference list.
3. Section 2.1, the basic information of the field experiment was missing. Readers can not get anything about the field experiment in such simple description now. Which time periods of the first and second data sets used for model validation? All of this information should be added in detailed.
4. Section 2.3, how many haze cases did the numerical experiments choose? I did not see any introduction about the time periods or haze cases through the manuscript. Or only one case form 27 December 2016 to 31 December 2016? Can the reliability of the results be confirmed by more haze cases? The introductions on sensitivity experiments were also missing.
5. Line 105, some studies had revealed that the turbulent characteristics of PM2.5 are different with heat, this information should be clarified.
6. Line 111, () was missing in “Noh et al., 2003”.
7. The parameterization of the new value of Kzmin was proposed abruptly without sufficient physical discussion. Line 129, the authors described “We assume that the value of EF can be used to characterize Kzmin in different regions.” Why did you propose this assumption? In other words, what is the physical meaning behind the formula 5? Why is it set up in this form? Line 223, “After setting the adjustment factor value to 1.0 in formula 3”, why did you set an adjustment factor if you “assume that the value of EF can be used to characterize Kzmin in different regions”? and What is the basis for setting 1.0 as the adjustment factor? Very confusing, was it formula 5 or 3 in Line 223? Under stable conditions, because of the high value of EF, the values of Kzmin were at least 100 times larger than under unstable conditions, Kh might be close to 2, was it reasonable? I highly doubt the physical rationality. Based on your formula 5, Kh under stable conditions might larger than  under unstable conditions. Anyway, the formula 5 you proposed was the core of this manuscript, more physical explanations are needed.
8. Line 134, “the value calculated by formula 1”, was it formula 5?
9. Line 140, the formulas of MB, IOA, RMSE, R, NMB and NME should be clarified at somewhere appropriate.
10. The mean model performance in Table 1 and Table S1 means the mean performance from several cases or one case in whole domain? Another key issue is that I did not see any comparison of the simulated and observed PM2.5 time series.
11. Line 238, “Figure 5d, 5e” means “Figure 6d, 6e”? same mistake in line 239, Figure 5 f.
12. Line 265, “Figure 7. The distribution of difference …”, you mean the PM2.5 concentrations difference?
13. Writing needs to be further improved.