This manuscript explores the sensitivity of Ångström exponent to aerosol radius in a CTM, and uses the information together with satellite data to design a hybrid AOD assimilation. The topic is interesting to the aerosol and satellite committee. The paper is well written and the method is clearly descripted.

My major concern is the authors attribute the discrepancy of Ångström exponent between satellite and model only to aerosol geometric mean radius. However, distribution of aerosol chemical compositions, aerosol mass density, refractive indices, relative humidity, hydroscopic growth, internal/external mixing in model assumptions all have uncertainties, some of them could be even more uncertain than aerosol size. Even we assume Ångström exponent is mostly sensitive to aerosol size distribution, the standard deviation of size distribution is also an important factor in additional to the radius. Are all these factors confirmed to have neglectable bias in the model, or their uncertainties hard to affect Ångström exponent? I would suggest add more analysis or at least discussion for this point.

Other comments:

- The posterior radius could be evaluated with more measurements to make them more convincible. For instance, a brief comparison with size distribution from surface or aircraft observations in the literature.  

- Any reason for the choices of wavelength pairs in Ångström exponent calculations (470-650nm for MODIS and 440-870nm for AERONET)? Would be the results different if choosing other wavelengths?   

- Section 4.2, paragraph 3. Modeled aerosol mass concentrations are underestimated. Depending on which chemical species are missing, this could largely affect Ångström exponent. Why do you think the bias of Ångström exponent is due to aerosol radius?

- Page 19, line 21. Please provide more evidence for assuming dust and sea salt are not important. What are the concentrations of dust and sea salt in your model simulation? Maybe this is also helpful to support your assumptions of their certain bins.   

- Fig 7-9: maybe narrow the range of colorbars to make the distribution clearer.

General comments:

This study attempts to investigate the effects of sequential assimilation of satellite-based aerosol size information (i.e., Ångström exponents) and aerosol optical depths (AOD) on the analyse of the aerosol concentrations. The assimilation experiments are conducted over the European region with the MODIS Deep Blue products. The results demonstrate that the assimilation of the MODIS observed aerosol size information could improve the surface fine particles analyses by correcting the model assumed aerosol geometric radius and subsequent the AOD observation operator. The paper is generally well written and scientific sound.

Main comments:

1. It looks the simulated Ångström exponents without any data assimilation are too low (below zero). The authours also claim that there are no dust events during the studied period, so does this mean that the default parameters of the aerosol radii for the fine aerosol particles such as the sulfate or carbonaces aerosol are too large in the LOTOS-EUROS model. If this is ture, why the model uses those values.
2. The results in Figure 2 demonstrate that there are some too low Ångström exponents. Probably, the quality of the satellite retrievals of the Ångström exponent over such region is not good. How does this unusual observation affect your assimilation result?
3. As the assumption of diagonal matrix of the model background covariance B for AOD, do you mean only the aerosol mass concentrations over the model grid with MODIS observation could be optimized? How about the model grids without any available observations to be assimilated? Does this induce some unreason aerosol distributions?
4. P19 Line 29, How to obtain the optimal aerosol radius using a 4DvEnvar? Please clarify it more detail?