This paper describes how the aerosol-radiation calculations in the WRF-chem model are modified to include more wavelengths that match with input needed by the model’s radiation parameterization. The subject matter is appropriate for GMD. While the paper is generally well written, easy to follow, and contains important information for users interested in aerosol-radiation interactions, there are several issues that need to be resolved before the paper is suitable for publication.

General comments:

Context: The paper demonstrates that differences between the “interpolated” and “resolved” simulations of aerosol optical properties and consequently aerosol radiative forcing can be large, usually for areas dominated by dust. However, they fail to mention other assumptions in the model which could impact those results.  Since the effects of dust in their “resolved” simulation seem to be related to absorption, model predictions will depend significantly on refractive indices used for both BC and dust.  Presumably dust in this model is assumed to be absorbing to some extent.  Yet, there are many studies measuring refractive indices of dust throughout the world, showing large difference resulting from different mineral compositions that vary from region to region.  Therefore, this parameter does not seem to be well constrained.  If one was to replace the refractive indices with less absorbing dust, the differences between the “interpolated” and “resolved” simulations would presumably be smaller.  In addition, the aerosol size distribution will have an impact on the aerosol optical properties, but the paper does not provide any information on how the size distribution of dust is represented. These are complicated issues which are coupled to model evaluation (see next comment).  At a minimum the authors need to include discussion somewhere in the text and the conclusions to put their results into the proper context.

Physical Interpretation: The paper focuses on differences between “interpolated” and “resolved” simulations, but there is very little investigation into the physical reasons for the differences (other than absorption in general) and explain that to the reader.  A little more discussion is needed. For example, the drop in SSA around 2500 corresponds to an increase in absorption.  Is the increase in absorption due to larger concentrations of coarse aerosols? How do we know whether the model not just overestimating dust?  While that might explain the different trends in SSA at longer wavelengths, it could impact the magnitude of the differences.  Or are there other factors?

Evaluation: Most of the figures show differences between the “interpolated” and “resolved” aerosol optical properties simulations.  While there are a few instances of comparing results with observations (e.g. AERONET in Fig 3, ERA5 in Supplemental material), there seems to be a missed opportunity to use more observations to convince the reader that the “resolved” method is indeed better.  For example, why not show comparisons with AERONET in the China domain?  Why not compare TOA forcing to CERES?  These comparisons are relatively easy to do.  There are also many observations of scattering and absorption that could have been used to quantify model performance and better understand how well absorption is handled by the new treatment.  Meteorological comparisons use ERA5, but ERA5 is a proxy for observations and might have larger uncertainties where in situ measurements are sparse, but the authors do not comment such caveats.  In summary, I would encourage the authors to do a more thorough job of evaluation.

Figures: The authors need to be more explicit in what their figures are showing.  I assume that most of them are monthly averages, but the figure captions and text need to make that clear.

Computational Aspects: The authors should describe the computational costs of computing the aerosol optical properties for the “revolved” vs “interpolated” method.  Presumably it is 3.5 times more, but they should be explicit.  Would also be useful to put those costs in perspective with the overall runtime of their simulations.  Somewhere in the paper (Section 2.3?), they should also note how often the aerosol optical properties are updated.  In addition, the aerosol optical property calculations in WRF-Chem work for buik, modal, and sectional representations of aerosol.  Do the modifications follow that format, or do they only work with the sectional representation used in this study?  While I do not think it is necessary to compare similar differences between the "interpolated" and "resolved" methods for bulk and modal aerosol representations it would be useful to let the reader know about these details.

Specific Comments:

Lines 37-40: These sentences were generally true several years ago, but many global models now can be run in a regional-refined mode similar to WRF or run globally at kilometer scale grid spacings.  While they may not have as sophisticated chemistry, they do have treatments of aerosols and aerosol-radiation interactions.

Line 69: Would be useful to include a table of physics parameterizations used in WRF, so users to not have to search for those in another paper.

Lines 88-89: Would be useful to number the equations.  Also, this method implies that AOD and forcing is linear.  Is that true?

Lines 135-142: It would be useful to include the AERONET site locations used in this study in Figure 1, rather than in S1.  Is there a reason other datasets, like MODIS satellite AOD, are not used?  It might be useful to explain to the reader that AERONET data is available at multiple wavelengths, but that may not be the case for other datasets.  I assume this is one reason, but there could be others.

Line 144-146: Somewhere you should mention that this result is averaged over space and time for the entire domain from the simulation. 

Line 148-151: This sentence has a lot of commas, so the thoughts are difficult to follow.  Suggesting breaking this up into at least 2 sentences so that the message is clearer.

Figure 3: is it necessary to go out to 3500 nm when the observations stop at 1500 nm?  T=I believe AERONET does have observations at these high wavelengths, but not for these stations?  Are the very long wavelengths important for the radiation calculations? If there is a reason to plot the long wavelengths, the text needs to describe why.

Line 164: It would be useful to include a figure similar to figure 3, but for the China domain.  Rather than showing results at individual sites, one could average the results over multiple sites.

Line 170: Can you provide an explanation for this dramatic decrease in SSA?  In addition, for the Sahara area, SSA from the resolved method is much smaller than from the interpolated method.  What explains this in terms of the aerosol size distribution?  For example, is there a lot of fine-mode dust?

Line 171-173: The authors describe how AAOD is derived.  Is this the same was as how AAOD is derived by AERONET?

Line 171: I think the reference should be to Figure 5 here.

Lines 206-207: The authors say that the TOA forcing is now more consistent with other studies for the Sahara.  While Y Feng (2023) does show positive values over the Sahara (their Fig 11), the maximum values are close to 5 W/m-2 (which is also similar to Albani et al. 2014).  Fig 7d has peak values that are 2-4 times larger.  So there is a qualitative agreement, but not a quantitative one.   Feng et al. (2022) has values > 10 W/m2.  Feng et al. (2022) seem to describe the implementation of dust in the MPAS model, using the same methodology as in WRF-Chem. Section 2.2.6 in that paper does not describe the use of the “resolved” algorithm, so I assume the “interpolated” algorithm is used.  But what is confusing is that MPAS produces a positive TOA forcing over the Sahara, but the WRF-chem simulation (Fig 7a) produces 0 to slightly negative values.  Why?  This statement seems to contradict what is stated in around line 206.  In addition, other WRF-Chem studies show positive TOA forcing over the Sahara (e.g., DOI:10.1038/s41598-020-69223-4) which is presumably using the “interpolated” method.  Not clear why different results are being obtained. It would be useful for authors to do more of a literature search on WRF-Chem used to predict radiative forcing over the Sahara.

Lines 233-237: I understand it is useful to look at surface temperature.  But 2-m observations are usually used to evaluate model predictions of temperature. I suspect that the impact on 2-m temperatures will be far less.

Line 237: What are the other factors?  Presumably altering aerosol-radiation interactions will also impact clouds which has direct effects and aerosol-cloud interactions could have other impacts as well.  They mention these effects for Figure 10, but doesn’t the same apply to Fig 9?

Line 255.  This statement is hard to judge based on Figure S2.  It would be better to show Figs S2c and S2f as “Resolved -ERA5” so that they can be directly compared to “Interpolated-ERA5”.  Also quoting some #’s on average bias over the domain (or subdomain) between the two simulations would be useful.

Line 266: I have the same comment for Figure S3 as Figure S2.

Lines 298-300: Since the differences over China are small, it would be useful to speculate that they would also be small in other areas where anthropogenic aerosols dominate?

The study presents the updated version of the aerosol radiative feedback algorithm in the coupled meteorology-chemistry model WRF-Chem v4.4. The manuscript also describes the current algorithm utilized in the default version of WRF-Chem. A full gas-aerosol chemistry scheme CBMZ-MOSAIC is used to simulate the aerosol radiation feedback for different regions of the world. One domain covers China, and another one covers the Saharan region. 

The model simulations with the default and updated algorithms show some differences in the aerosol optical depth and single scattering albedo, in particular for the Saharan region. The authors show better agreement of the model with the AERONET AOD data when the updated aerosol radiative feedback algorithm is used.

The updated algorithm can help to improve the WRF-Chem air quality and meteorological simulations. I recommend publishing the manuscript after addressing the following comments:

• Lines 235-240: You mention the indirect feedback. Are you referring to semi-direct feedback? I assume your WRF-Chem configuration does not include the aerosol-cloud interactions.
• The resolution of your WRF-Chem grid is quite coarse (50km). I’m wondering if high-resolution simulations (e.g. 4-10km) would yield somewhat different results. For instance, anthropogenic pollution can be better simulated by a high-resolution model.    
• I recommend adding the satellite AOD data for the evaluation of the model sensitivity simulations. The spatial coverage of the satellite observations would greatly complement the sparse AERONET observations used in the model evaluations. 
• Given the significant impact of the updated algorithm on heating rates over Sahara, it’d be helpful to evaluate the temperature simulations by WRF-Chem. Such an evaluation would help to determine whether the improved AOD and SSA simulations lead to improvement of the meteorological simulations. The authors present some comparisons with the reanalysis data, but comparing directly to the surface stations (e.g. 2m air temperature) would be really informative. To my knowledge, some global models such as GFS don't assimilate the ground-based weather observations.