The manuscript documents the exceptional aerosol conditions over Athens during the summer 2023 Greek wildfire season, focusing on two major episodes (18–21 July and 22–25 August). It combines MODIS fire detections, FLEXPART retroplume simulations, multi-wavelength Raman/polarization lidar, AERONET sun–sky–lunar photometry, in-situ PM/BC and ACSM chemistry, and Brewer UV spectral measurements to characterize aerosol vertical structure, optical/microphysical properties, and the associated impact on surface UV at 320 nm (treated as a proxy for UV-B). Reported conditions include AOD up to ~1.2, high fine-mode fractions, SSA spanning ~0.85–0.98, PLDR variability consistent with smoke–dust mixing, and surface 320-nm irradiance reductions up to ~50% during peak smoke.

My recommendation is major revision. The dataset and the synergistic concept are strong, but several central methodological choices (especially the UV/SSA retrieval setup, cloud screening, and uncertainty propagation across the chain) are not yet documented or stress-tested enough to support some of the stronger interpretations.

1. The paper equates 320-nm irradiance with “UV-B impact” but does not demonstrate how well 320 nm tracks band-integrated UV-B (290–315 nm) variability under changing ozone and aerosol conditions.
2. The UV closure relies on cloud-free simulations, yet the lidar time–height plots clearly show cloud formation during key windows; without explicit cloud/thin-cloud screening rules and the retained time stamps, the UV comparisons are hard to trust.
3. The SSA(320) retrieval uncertainty is treated mainly via ±5% irradiance bounds, but AOD(320), surface albedo, aerosol vertical distribution, and phase function (dust vs smoke) can easily dominate the error budget; some structured sensitivity tests (or an error propagation table) are needed.
4. AOD extrapolation to 320 nm using Ångström behavior (e.g., 340–440 nm) can be biased in coarse-mode/dust or mixed cases (spectral curvature); the “~0.05 agreement” should be broken down by event type (smoke / dust / mixed), not just pooled statistics.
5. There is an internal tension that needs clarification: the manuscript states the least UV attenuation occurs on the dust day (23 July), while also retrieving very low SSA(320) ~0.75–0.8 for dust; it is not clear whether the comparison is for global vs direct vs diffuse irradiance and whether residual cloud effects or geometry explain this.
6. The microphysical inversion assumes a wavelength-independent complex refractive index across 355–1064 nm; this is a strong assumption for smoke (possible brown carbon) and dust, and its effect on retrieved m_i, R_eff, etc. should be quantified.
7. Treating particles as spheres for PLDR<10% and spheroids otherwise is too binary for mixed layers; at minimum, show threshold sensitivity (e.g., 5/10/15%) or discuss a continuous mixing approach.
8. POLIPHON mass profiles depend on fixed parameters (e.g., smoke density 1.35 g cm-³, variability allowances, “non-collocated still effective”); this may be reasonable for stable dust, but wildfire plumes are temporally/vertically variable—please justify applicability for the smoke cases and propagate these choices into uncertainty bars.
9. FLEXPART configuration needs better alignment with observations: if aerosol layers reach ~5–6 km but the release layer is limited to 0–4 km, source attribution may be incomplete; sensitivity to release height and back-trajectory duration should be demonstrated.
10. The smoke/dust/mixed attribution is mostly qualitative across multiple indicators (AE, FMF, PLDR, LR); a single reproducible classification rule (even simple thresholds with rationale) would make the “synergy” claim much stronger.
11. Lidar-retrieved and AERONET-retrieved effective radius differ substantially; the discussion currently stops at “different sensitivities,” but the paper should compare layer-integrated vs column quantities and explain whether the discrepancy affects downstream radiative interpretation.
12. Statements about “strong absorption” lean heavily on retrieved m_i and UV-SSA; please tighten the cross-validation with independent constraints (AERONET SSA variability, eBC source apportionment, and ACSM organics) so the absorption narrative is not driven by one retrieval stream.

The manuscript links elevated free-tropospheric smoke layers (2–5 km) with surface PM/chemistry peaks, but the coupling mechanism is not demonstrated; a short analysis using BLH, timing consistency, and evidence for downward mixing would help separate transported impacts from local/BL contributions.