The paper "Exceptional high AOD over Svalbard in Summer 2019: A multi-instrumental approach" by Herrero-Anta et al. is a thorough study of an episode of enhanced aerosol at Svalbard during summer 2019. It combines observations of several instruments to identify aerosol characteristics. Further, modeling with FLEXPART is used to identify the different sources of the aerosol.

Overall, the paper presents a comprehensive, logically structured study, it is well written, and is of interest to the broad readership of ACP.

The paper is therefore recommended for publication in ACP after addressing my minor comments as detailed below.

My main comment is that in some places discussion of some features in the data is missing.

SPECIFIC (MINOR) COMMENTS:

(1) Fig.2: Please comment!

Why is there a peak of AOD with strong standard deviation in the reference record between 1 and 15 July?

Is this a repetitive event each year?

Or is this peak attributable to a specific event? If yes, which?

(2) Fig.4: For the S1 event, beta_sca is only enhanced at GAL, but not at ZEP, while beta_abs is enhanced at both sites.

Do you have any explanation for this?

(3) l.301-311, about Fig.5: ZEP discussion is missing!

Here you should also comment about the PNSD at ZEP, which does not show a clear bimodal structure during the surface events, and the distributions peak at sizes between Aitken and Accumulation mode.

(4) l.324-325: What about the other layers seen by KARL?

There is a more intense layer at 13km that is not seen by CALIOP. Why?

Is this an issue of the CALIOP sensitivity?

(5) l.447-449: Thin cirrus clouds are hard to detect by ground based and space based instrumentation. Could thin (subvisible) cirrus clouds also contribute to negative values of delta-DNI?

TECHNICAL COMMENTS:

(1) l.28: levels.Lisok -> levels. Lisok

(2) l.52: Siberian wildfires -> smoke of the Siberian wildfires

(3) Table 1: abbreviations of several parameters (e.g., DNI, DIF) are only given later in the text. This should be mentioned in the table caption.

(4) l.148: can be also be -> can also be

(5) l.161:

to Reference Upper-Air Network (GRUAN).

->

to the Global Climate Observing System (GCOS) Reference Upper-Air Network (GRUAN).

(6) l.171: from zero to the unit, being small -> from zero to unity, with small

(7) caption of Fig.2, l.2-3: Sentence "Long-term daily means ..." can be deleted because same info is given at the end of the caption.

(8) caption of Fig.2, l.5: with errors bar -> with error bars

(9) l.253: one maxima -> one maximum

(10), (11) l.254: main maxima -> main maximum

second maxima -> second maximum

(12) l.259: This longer radii -> These larger radii

(13) p.11, last line: longer -> larger

(14) l.277: next mean values: -> following mean values:

(15) Table 3 and text on p.13:

Here you use B_abs and B_sca instead of beta_abs and beta_sca.

Please use consistent notation throughout!

(16) caption of Fig.5: the are only -> there are only

(17) l.303: With respect GAL observations -> Regarding GAL observations

(18) l.331: is shown -> are shown

(19) l.361: row)is -> row) is

(20) Caption of Fig.10: red line -> magenta line

(21) Caption of Fig.11: With respect the sources -> With respect to the sources

(22) Caption of Fig.12: with respect the reference -> with respect to the reference

(23) l.495: Ship born -> Ship borne

(24) l.503: under reques to the authors -> under request to the authors.

General comment: The authors analysed Arctic aerosol observations and discuss the findings. Pollution source identification based on FLEXPART modelling is part of the study. The focus is on the summer of 2019. Wildfire smoke, anthropogenic pollution as well as volcanic sulfate aerosol (originating from the Raikoke eruption) polluted the troposphere and lower stratosphere from the surface up to about 20km height. The manuscript contains interesting information and is clearly worthwhile to be published in ACP. However, many questions came up during reading and need to be clarified as part of the revision of the paper. Major revisions are needed.

Detailed comments and questions:

Line 9: please state clearly: do you mean diameter of radius? …. 0.1-0.2 micrometer. Accumulation mode particles cover the radius size spectrum from about 100 to 500 or even 1000 nm! What do you mean with 0.1-0.2 micrometer?

Line 23: please be more precise: Do you mean the boundary layer or the free troposphere. Aerosols in the free troposphere are usually related to long-range transport, and not local aerosol production.

Lines 25-28: Besides the given references one needs to mention recent observations from MOSAiC (Ohneiser et al., ACP, 2021, Ansmann et al., ACP 2023) and also the satellite observations presented by Kloss et al. (ACP, 2021).     

Line 30: What about indirect aerosol effects, i.e., the impact on water cloud, mixed-phase cloud and cirrus formation, and related precipitation processes.

Line 51: Please use lofting instead of lifting throughout the article!

Line 53-55: The arguments show already that in situ observations at ground are not just helpful in the study of the aerosol conditions in the entire vertical column. Especially removal processes by washout events permanently clean the lowermost 200 m of the Arctic troposphere so that surface observations cannot be used to describe the cloud- and radiation-relevant aerosol conditions in the Arctic. Such statements should be included in the article. Furthermore, how is the 2% contribution by biomass burning identified? If this finding is based on BC information, the conclusion may be wrong. Wildfire smoke consists to 95-98% of organic carbon (OC).

Line 81: mention season and year of the R/V OCEANIA field studies!

Line 135: KARL seems to be a very powerful lidar (50 laser pulses per second, about 200mJ per pulse at 355, 532, 1064nm, 70 cm telescope)! Why are almost no lidar observations shown? I expected particle extinction and smoke lidar ratio profiles at several wavelengths! Some Raman lidar applications! But only a few low-quality color plots are presented together with not trustworthy inversion products without showing any basic multiwavelength lidar observations. This is not good, and should be improved! I will come back to this point in more detail later on in this review.

How are the aerosol backscatter coefficients computed? I assume by using the Fernald method! What particle lidar ratios are assumed in the Fernald data analysis at the different wavelengths? In the case of aged wildfire smoke, the lidar ratios are about 55 sr (at 355nm), 85sr (at 532 nm) according to the report of Ohneiser et al. (2021)  and about 100 sr (at 1064nm) for aged smoke as described in other papers. Is such a lidar ratio spectrum considered? In the case of sulfate aerosol (Raikoke) a similar lidar ratio spectrum holds but with lower lidar ratios, maybe 35, 60, and 75 sr. All this needs to be mentioned. The basic lidar products are the backscatter coefficient spectra and they are influenced by the lidar ratio input values.

Is the use of simply three backscatter coefficients really sufficient to retrieve the effective radius of the particles? I would not trust these results, especially not in the case of different aerosol types above each other which may be even partly mixed in the UTLS height range.

Lines 185-189: CALIOP extinction profiles are used. When using the CALIOP elastic-backscatter lidar profiles, again a lidar ratio has to be assumed to obtain the backscatter and extinction profiles! I guess, the CALIOP science team used 70 sr for smoke and about 40 sr for sulfate particles. Please provide numbers here. The uncertainty in the products are high, higher than 50% (in terms of relative errors), I speculate!

Line 201: To my opinion, more information (some kind of a general overview and introduction) on the optical properties of smoke particles is needed. The BC content of aged smoke is 2-3% as can be found meanwhile in many modelling papers (references may be found in Ohneiser et al., 2023). Smoke particles mainly consist of OC (95-98%). Particle density of smoke particles is roughly 0.9-1.3 g/cm3. The OC content contributes to self-lofting  because of the ability to significantly absorb even at wavelengths greater than 400 to 500 nm. In the case of a smoke AOD of 2-3 at 500 nm, self-lofting leads to ascent rates of 3 km per day. Ascents rates of 500 m per day are still possible for AODs of the order of 0.5-1. In the case of 30 days of transport, pronounced smoke layers may thus ascent, on average by 100 to 200 m per day, and in the beginning (shortly after emission) when the smoke plumes are optically dense, on average by 500 to 1000 m per day. Even if self-lofting is not considered in the FLEXPART simulations, a discussion on the consequences is needed. All in all, section 2.2.6 must be updated by considering self-lofting aspects.

Line 235: Why do you not use the period from 2002-2018 as reference?

Line245: The Siberian fire episode from the beginning  of July 2019 to mid of August 2019 was already discussed in Ohneiser et al. (ACP, 2021 and 2023).

Line 251: … when the peaks oscillate from 0.01 to 0.04 …. Please explain precisely: what do you mean here? 

Line 269: How trustworthy are the SSA values? The MOSAiC multiwavelength lidar observations also show smoke SSA values of 0.95-0.96 (Ohneiser et al., 2021). Could be  mentioned as a support.

Table 2: The figure caption should mention that the products are derived from photometer observations.

Line 275: What do you expect from the surface observations? Washout processes continuously remove particles entering the PBL from above. Smoke plumes travel in the free troposphere and in the ULTS height region. Raikoke aerosol travels at stratospheric heights. So, what can tell us the Arctic surface observations about smoke and volcanic aerosols in the Arctic?

Line 316 (page 14): Stratospheric AOD values as presented in Figure 6 should also be discussed in the main text body and contrasted to the volcanic AODs. The volcanic AODs are probably at all smaller than 0.025 with occasional exceptions, but at all below 0.05.

Line 321: Only four KARL (ground-based lidar)  observations in 4 months (120 days)! This is really bad news!

Figure 6: This figure showing CALIPSO profiles is also of low quality. Many height bins are zero! How can we trust  such extinction profiles? What are the numbers for the extinction peaks? There is no x-axis for extinction values. Again, what particle lidar ratio is used in the CALIOP data analysis? Many peaks are above the tropopause, some are caused by smoke (then a lidar ratio of 70 sr is appropriate) and some by volcanic sulfate layers (then a lidar ratio of 40 sr should be used in the Fernald retrieval). Any comment on this issue is welcome!

Figure 7: What is shown? Is the backscatter signal profile shown? The color scale indicates: the ratio of the aerosol backscatter to the Rayleigh backscatter coefficient is shown. Please clarify!

I do not see any consistency between Figure 6 (always sharp layers with the vertical thickness of less than 1 km) and Figure 7 (vertically deep layers, partly 5-7 km thick, most layers without sharp edges).

Why do you not show any figure with all the basic lidar profiles, i.e., height profiles of the backscatter coefficient an 355, 532, and 1064 nm together with the particle depolarization ratio. This would be a convincing figure as an introduction to  Figure 8!

Figure 8 shows some kind of inversion products (effective radius estimates). How can I trust Figure 8? Without showing any figure with the input profiles for the effective radius retrieval, I have to conclude that these backscatter profiles are of rather low quality so that the retrieved effective radius values are also of rather low quality (and therefore not trustworthy).

As mentioned above: Each backscatter computation at 355, 532, and 1064 nm needs quite very different lidar ratios as input! And these lidar ratio spectra are quite different for smoke and for sulfate. The input lidar ratio sets have a big impact on the quality of the determined spectrum of the backscatter coefficients, and consequently, on the effective radius values. Again: How trustworthy are the computed backscatter ratios and at the end the estimated effective radius values? All this must be discussed in the paper.

Ohneiser et al. (ACP; 2021) show effective radii for the Siberian smoke of 0.2-0.22 micrometer. Since the size distribution of volcanic sulfate particles is similar (well-defined accumulation mode) similar effective radii are expected for the Raikoke particles.

Lines 332-355: The discussion on page 16 and 17 is very speculative. Speculations should be avoided as much as possible. The effective radius values shown Figure 8 for the troposphere are confusing. It is impossible to determine the effective radius for both the fine mode and for the coarse mode from just three backscatter coefficients. 

Line 365: A discussion on self-lofting of smoke layers needs to be included. As mentioned  aged smoke particles (i.e., smoke older than 2-3 days) consists to 2-3% of BC and 97% of OC. For these BC-OC particles, lofting efficiencies must be estimated (Ohneiser et al., ACP 2023). One may show Figure 9, but needs to discuss the potential impact of lofting that shifts the profiles upward, towards greater heights. In this discussion, the findings of Ohneiser et al. (ACP, 2021), partly summarized in Ansmann et al. (JGR, 2024), maybe helpful. The MOSAiC observations show the aerosol pollution conditions in the Arctic for October-November 2019. Self-lofting probably came to an end in September 2019, before the MOSAiC expedition started. 

Lines 395-399: The maximum conversion of SO2 into sulfate occurs about 6 weeks after a volcanic eruption. For Raikoke (22 June eruption) the maximum sulfate load should have been observed around 10 August. However, freshly formed volcanic aerosol layers are usually organized  in sharp layers in the stratosphere as indicated by the CALIOP observations in Figure 6. The thick layer from 7 to 15 km is, to my opinion, a composite of smoke layering with sulfate layer contributions on top (in the stratosphere). Karl observations in Fig. 7 seem to be in line with this hypothesis. However, it cannot be excluded that some dense smoke layers also entered the lower stratosphere (by self-lofting). Thus, the interpretation of the observations needs to be carefully done. The observations in Europe (Vaughan et al., 2021) have to be handled with caution as well. It remains open to what extent smoke and sulfate  contributed to the observed aerosol pollution in the stratosphere over the UK.

The conclusion section as well as the Abstract need to be updated after the revision of the main parts of the manuscript.

Lines 493-500: Dedicated (future) field campaigns make sense, but all the methods, techniques and instruments are already available since decades, but well designed actions have never been conducted in the Arctic.

Figure 9: AFR in the panel, AF in the caption.