The manuscript presents the algorithm for analysis of aerosol mixture composition, based on expected EarthCare observations. Authors consider 4 main aerosol components (two types of fine and two types of coarse mode particles) and try to present the observed particle volume as a sum of these components volume, using OEM approach. Atmospheric aerosol is a complicated object and particle parameters are strongly variable, still this looks like a reasonable approach, because number of independent observations from EarthCare will be very limited. Manuscript is well written, contains new important results and is suitable for AMT. For me, personally, was interesting to see, how additional 532 nm measurements will influence the results obtained from UV channels.

Technical comments.

p.6. Choice of particle model parameters is always an issue. In particular, parameters depend on RH. This especially critical for maritime aerosol and for small non-absorbing particles. Probably authors could mention range of RH, where their model is applicable.

Table.2. Authors assume that depolarization ratio of small particles is 3%. In practices, however, we often observe smoke and sulfate particles with depolarization up to ~8% or even higher. Small values of assumed depolarization ratio probably may increase contribution of dust particles in the examples provided in Fig.11.

For smoke particles the lidar ratio at 532 nm is usually significantly larger than at 355 nm. Can it be obtained from component parameters in Tab.2? The same question is for dust. In experiment lidar ratio of dust at 355 sometimes significantly higher than at 532.

Ln.231. As I understand, sometimes contribution of a single component may exceed 1.0. Just wonder, if it is possible to add additional constraint, that the sum of all components must be 1.0?

Ln.238.”… if the total relative volume contribution (sum of the relative volume contribution per aerosol component) is greater than 1, then the state vector is normalized”

This is actually related to my previous question.

Ln.239. “…There is no constraint in place  in the case of a total relative volume contribution that is less than 1.”

Why no normalization this case?

Fig.5. Probably no reason to show lidar ratio and extinction below 1 km.

The paper by Floutsi et al. introduces a novel aerosol classification method with high flexibility in terms of channel configuration. The output considers aerosol mixtures of four aerosol components that are directly linked to the lidar intensive properties. The authors provide the necessary information and guidance to understand the optimal estimation method, and, furthermore, to perceive the methodological ramifications through suitable visualization and specific case studies. The language and presentation are commendable throughout. The paper therefore is suitable for publication with only typographical corrections.

In the following, I included only a few minor comments and corrections that I hope will improve the manuscript.

Minor Comments:

Ln117: I suppose that 30 iterations is an empirical estimate. How long does it take for the algorithm to provide output? Can it be used operationally?

Ln144-148: Do you apply quality screening to the input optical parameters? If yes, what are the threshold of acceptance for these parameters? Also, what does “order of appearance in the vector” mean ? Does it imply the discriminatory power of the intensive properties?

Ln236-239: Why the penalty term is not enough? Is the second criterion only invoked when the total relative contribution is greater than 1?

Ln320: It would be better to stress earlier that the input is layer-averaged values. Could it be possible to apply HETEAC-Flex to high temporal resolution lidar maps?

Figure 7 and Ln370-371: Why color ratio and particle depolarization ratio do not have error bars? This is a bit confusing for me. Also, do you consider the standard deviation of the averaged intensive profiles for the error estimation? 

Ln354: Can you expand on how you defined the aerosol layers? Is layer detection part of the methodology?

Corrections:

Ln79: Consider changing “reveal”.

Ln92: Remove “aerosol typing scheme”.

Figure 1: Add “the” before “HETEAC-Flex”.

Ln98: Consider removing the hyphen.

Ln131-135: Please split this long sentence.

Equation 9: One of the brackets is not closed and is not needed.

Ln250: Rearrange “linear particle”.

The paper by Floutsi et al. introduces a novel aerosol classification method based on lidar derived intensive properties. The main results of the study are of interest. However, authors should point out the novelty of their study also regarding the existing typing schemes. I recommend the publication of the manuscript after minor revisions, considering some general and specific issues detailed below in my review.

General comment

Why did authors not mention all the existing typing schemes based on lidar intensive parameters? The added value of their aerosol classification method should be pointed out.

Line 112. Authors state that: «the covariance matrix S_ describes the measurement errors». How are measurement errors are being calculated?

Line 117. Authors state that: «Typically, the process converges within 30 iterations and if not, then it fails to converge and, consequently, there is no optimal solution.» How this value is being selected?

§3 Application of HETEAC-Flex. How are layers been defined? Authors should provide information on the layering detection.

Figure 7. Errors should be added in all products.

§3.3 has different structure than the 3.1 (Case, Overview and Aerosol characterization) and 3.2 (Case, Overview and Aerosol characterization). Paragraph 3.3 should be homogenized with the previous ones.

Figure 6. Why are these retrievals modes selected? Authors could present all the available modes for this specific case study.

Specific comments

Figure 8. Legend shouldn’t be color filled.