Review on paper titled “The impact of assimilating Aeolus wind data on regional Aeolian dust model simulations using WRF-Chem“ by Pantelis Kiriakidis et al.

Paper compares output from two WRF-Chem model runs: with and without assimilated wind fields. In particular, authors use extensive set of ground and satellite observations to estimate model skill to simulate dust events. The article is well written and easy to follow. Authors conclude that they got significant improvements in dust simulation over the EMME region using assimilated wind data, while comparisons with the whole simulated domain diffused the improvements. 

My main concern is the ability of WRF-Chem to correctly simulate the dust cycle in this study. Fig. 2 demonstrates it, i.e there is a better agreement in PM10 between runs rather than between runs and observations. See also Fig. A2, where good agreement between models is shown. Model also overpredicts PM10 and not capable to capture high pollution events in most of the cases. Thus, I think, authors pay attention to the 2nd order effect, while 1st order effect (dust simulation itself) is not satisfactory resolved. Therefore, I recommend revision before accepting for publication.

General comments:

the title is a misleading. It is not clear, whether WRF model itself assimilates wind data or not.

Introduction is lengthy, 2nd paragraph on page 1 and 1st on page 3 could be shortened.

References [2] and [3] devoted to rigorous dust simulation on the Middle East are missing in the manuscript.

Not clear why authors used complex (MADE/SORGAM) aerosol scheme to simulate dust? Is there any justification for it? However, here [1] you may find some useful details on how to simulate dust in WRF-Chem using modal aerosol scheme.

Specific comments and technical corrections:

Line 62: first occurrence of WRF-Chem. Please add reference.

Line 90: HSRL, HLOS unknown abbreviations.

Line 107: .. seasons of the region.  Please specify which region.

Line 130: Natural emissions. Please explain, what do you mean?

Line 170-172: what type of FDDA you used? Not clear, who lateral boundary conditions can be improved by FDDA? If FDDA is enabled in WRF, then model fields (not observations) are nudged to reanalysis fields.

Line 188: remove ; ?

Line 200: height of (t,V,lat,lon) model level, and ΔH - width of the (t,V,lat,lon) model level.

Lines 200-201: Please remove in in units.

Formula 1: add (t,V,lat,lon) to PH and PHB

Formula 2: Replace by AOD(t,lat,lon)=∑EC55(t,V,lat,lon) ∗ ∆H(t,V,lat,lon), where ∑ - sum over V.

Figure 1: Land contours are hardly seen (Fig. A3 same).

Line 301: missing formula for IOA.

Line 317: 14-19th. Please add October.

Figure 5: what os hel1, 4 on plot legend? Please cut top altitude to 5km.

Figure 6: Please replace AERONET-alpha by “Ångström Coefficient”, replace y-axis label by “Ångström Coefficient”

Figure 8: Please replace ‘black boxes’ by ‘black rectangles’.

Figure A1: Please move it to the main text and illustrate all (if possible) geographical locations mentioned in the study. Also plot locations of AERONET stations (remove Fig. 4a, which is empty anyway).

References:

1. Osipov et al, Severe atmospheric pollution in the Middle East is attributable to anthropogenic sources
2. Parajuli et al, Dust Emission Modeling Using a New High-Resolution Dust Source Function in WRF-Chem With Implications for Air Quality
3. Ukhov, A. et al. Assessment of natural and anthropogenic aerosol air pollution in the Middle East using MERRA-2, CAMS data assimilation products, and high-resolution WRF-Chem model simulations

General comments:

This study aims to demonstrate that the assimilation of Aeolus in the model used as boundary conditions for the dust simulations improves the capability of the regional model to resolve dust loads.

The paper is readable and the results are validated from external measurements.

The study is based on several months of simulation, but only focuses on a reduced period and one particular event. Studies with Aeolus data on other specific events (e.g. tropical cyclones) acknowledge that it is difficult to work on such a reduced set of events (see for instance DOI: 10.1002/qj.4370, The characterization and impact of Aeolus wind profile observations in NOAA’s regional tropical cyclone model(HWRF) by Marinescu et al., 2022). The impact is not systematically in the direction of the average impact (i.e. assimilating Aeolus data can also be very detrimental, in some cases).

I think this paper requires such a discussion on the significance of the results when considering the reduced set of events.

Without this discussion, this paper is still an important contribution to the demonstration of the usefulness of Aeolus data, through a well-documented case study. However, I don’t think it stands as a solid proof by itself.

Specific comments:

I do not understand the expression “comparisons with the whole simulated domain diffused the improvements” (l. 355) or “Statistical comparison of all 56 AERONET stations within the extended model domain diffuses the improvement” (l.436) and another occurrence on l. 468. In particular, this use of the word “diffuse”. This might be a specific jargon that I am not aware of, but could you explain this notion with a different wording?

l.88 Is “turmoiled” necessary?

l.109: Did you mean “in section 2.1 to 2.5”?

l.197 “miscrophysical”

l.260 “deterrent”, not sure I understand, did you mean “inherent”?

l.263: why was the nearest hour so different from a 3-hour average? Is the model AOD noisy or extremely variable?

l.265-269: that’s a lot of averaging. How different are the values where there is an overlap for instance?

In addition, I don’t understand what is being produced here. Maps of MODIS AOD?

l.295: Please describe better figure A2: It shows that both model runs give very similar results at the Agia Marina station during the spring period.

l.305-306: statistical significance again… This contradicts the l.308 statement of a “thorough investigation”

Fig. 5: I would suggest to either remove hel1 and hel4 from the legend or introduce it somewhere in the text.

l.334: Isn’t it 4 FLEXPART runs but only “two, 5-day periods”?

l.342: Fig A4 does not show the AERONET stations

l.411: LIVAS is a dataset, not a “lidar”

Fig 8: Is it possible that dust events happening close to the domain boundary are less well resolved? (e.g. dust could be transported from outside the domain, across the boundary). There are also some discrepancies to the East of the Caspian Sea for instance. The other hypothesis would be that getting the magnitude wrong on strong events already produces a large error. And unfortunately, they happen close to the domain border for this period.

General comments

The authors have studied the impact of using Aeolus-improved meteorology initial and boundary conditions in WRF-Chem regional dust simulations. More specifically, the initial and boundary conditions are IFS fields obtained with the assimilation of Aeolus wind profiles that go to improve meteorological patterns and therefore dust transport. The paper is well written and structured with a clear scope and meaningful results. I have some some questions for the authors and also comments that could help improving the paper.

Specific comments

- I find the title misleading. It reads as if Aeolus wind data were assimilated in WRF-Chem. This is not the case. These data are assimilated in the IFS and IFS outputs are used as initial and boundary conditions of WRF-Chem. I strongly suggest to change the title to reflect more faithfully the work done.

- similarly, I suggest to rephrase the many sentences in the paper that describe the experiments done as data assimilation experiments, which, in my opinion, is not strictly correct (and unfair toward the amount of work that went into assimilating Aeolus in the IFS), unless I’m misunderstanding the simulations. If the latter, it would be good to clarify how the Aelous data are used in WRF. In my understanding IFS analyses of wind, temperature and moisture (with or without the assimilation of Aeolus wind profiles in the IFS) are used as boundary and initial conditions of WRF through nudging.

- it would be good to provide more details of the nudging performed: frequency of nudging, nudging time scale, ramping period (if those apply).

- please clarify the calculation of aerosol optical depth and the definition of some of the variables used. “total-column atmospheric extinction coefficient at the wavelength of 550 nm (EC55)”: is not a profile instead of being total column (in eq 1 has in fact the dimension vertical layer)? “EC55 can be used as a proxy for dust optical depth”: not really a “proxy”, one is a profile, if I’m correct, the other is column-integrated, one refers to all types of aerosol and the other only to dust. Also, please review eq 1 (it is a line of code that must have been inside loops) so that is expressed well mathematically. Also, which are the optical properties and assumptions on particle size and shape used to derive EC55 from mass concentrations?

- the flexpart part could be removed. What does add to the study?

- it seems that the main impact you see is in the transport of dust, why do you think you did not see a greater impact on the mobilization and emission fluxes? could you comment more on this aspect please.

- sec 2.1 should not the scaling proportionality constant for the dust emission flux be unitless?

- sec 2.2 please review the paragraph, overall it does not read too well. For example, talking here about the “the control and assimilated runs” might confuse the reader since the same names are used for the WRF simulations; “2B10 baseline” maybe not all the readers would be familiar with this; “Rennie (2021) who provided the configuration” does not read well. Also, could be useful to specify more details: which IFS output is used by WRF, analyses or forecasts, and at which spatial and temporal resolution.

- throughout the paper there is a bit of inconsistency among the length of the experiments. It would be good to clarify the dates. In the introduction: “April - May and September – November 2020” in 3.1: “was run for two months in spring and two in autumn”. In 2.3: “periods 2020/03/04 - 2020/05/31 and 2020/09/01 – 2020/11/04”. Btw, why March is not analyzed? Also, is a spin-up period taken into account for both seasons?

- sec 2.5.4: why both MIDAS and WRF-Chem AOD have been regridded rather than regridding only the finer resolution one? I’m not suggesting to redo things, just to understand the reason for a final overall coarser spatial resolution of 0.4.

- line 263: “the latter was performing poorly” relatively to what metrics?

- Figure 4 does not contain the stations. Additionally, please consider reporting here or in A1 the main geographical names of places used in the discussions throughout the paper, and also the lidar station location.

- sec 3.2 Not clear why the AERONET AODs are filtered for dust-dominated conditions when the model output is AOD and not DOD (line 248: “current WRF model version does not output DOD”)

- Fig 6: is only AE>=0.75 shown? Can you please state it clearly. Also, the yellow color is difficult to see and to distinguish from the red.

- 368: the plots show high concentrations also lower than that, at 1 km. Similarly line 410, I don’t see the statement being consistent with the plot.

- 388-390 where is all this about AE shown?

- Table 3: define somewhere the regions considered here. Line 423 says that no sub-regions are considered, but there is one considered beside the whole domain.

- Lastly, it would be fair to add in the conclusion that the results are relative to a case study only, basically a part of the month of October makes the main difference, and that they are not necessarily statistically significant

Technical corrections

- 24: disease → diseases

- 85: please review the whole sentence: “an observational coverage network able to feed the model” does not make proper sense as it is written.

- 88 combated → overcome?

- 104: “the ECMWF-IFS assimilated Aeolus wind fields provided by ESA are implemented in the WRF-Chem” is not written correctly, please rephrase it. ESA provides the wind fields, these are assimilated in the IFS and then used in WRF-Chem.

- 106: in consideration → in the inclusion or not

- 109: remove or substitute “Following”, also in line 151

- 114: “the incorporation of the Aeolus assimilated wind fields within the ECMWF-IFS datasets” could be phrase it better. The wind fields are assimilated in the IFS model.

- 120: “adjusted”: in which respect?

- 122 mean bias does not need to be capitalized

- 126 EDGAR-HTAP: use first the full explicit name and then the acronym. Please check also other similar occurrences.

- 136: please check the grammar or the word “treatment” here: “was consistent with [ … ] aerosol size treatments”

- 140: “more accurate”: please specify more than what.

- 178: “the assimilated ECMWF-IFS dataset” is not strictly “assimilated”

- 194: “measure”: they are rather retrievals from measurements

- 200: “total”?

- 237: “model vertical bins”?

- 286: Oct. → in October

- 291: specify that these statements are for the pressure level considered.

- 296: “could be confirmed from long-term runs” → should be confirmed by

- Figure 1: Spring and Autumn don’t need to be capitalized, caption and title, please check the rest of the paper

- Figure 1 in average → in monthly average

- Figure 1: “and wind vector differences” → and of wind vectors (are not differences)

- Figure 1: please specify the months too, left and right.

- 302 “range 0” –> range from 0

- Figure 2: trendline? Please remove it.

- 353: rephrase it please. It is not the “comparison” that reduces the improvement, but is the improvement that is less strong when considering all stations.

- 370 attribute to→ explained by

- 374 SEVIRI → SEVERI as shown in Fig …

- 404: inactivity → activity

- in 3.3: locality → location

- 417-423: please review it, it does not read too well

- 424: instruments → datasets

- 446: report → paper

- 449 was more profund → was only?

- 451: Meanwhile?

- Fig 2: too low resolution.