Review of Analysis of a saline dust storm from the Aralkum Desert - Part 1: Consistency of multisensor satellite aerosol products by Xin Xi

This paper describes a dust outbreak over the Aralkum desert, claiming to address a knowledge gap in the representation of satellite observations of saline dust emissions. Several established aerosol products, like UVAI and AOD from NASA and NOAA sensors are compared, along with somewhat newer aerosol products like MAIAC AOD and aerosol optical centroid height (AOCH) from EPIC, compared with AOCH from CALIOP.

The first part is extensive and reads like a review of current satellite remote sensing techniques, but focuses only on NASA/NOAA single view instruments with their known limitations, which are confirmed in this study. In section 2 a review is given, suggesting a comprehensive description of Aerosol Remote Sensing of Desert Areas. However, the description is general in the sense that much is touched upon, but also leaves out many details for a full review, so that it is unclear if it is meant to be specific for this case, or a complete description for the interested reader. E.g. the UVAI description repeats information from the late 90s and describes no new developments. The different algorithms, cloud-corrected or not, is not described. It claims that 'historically, two wavelength pairs have been used", but does not mention the reason of channel availability for the 340/380 nm choice (and some other wavelength pairs not mentioned) and the important more optimal choice of 354/388 nm outside ozone absorption bands that can be implemented for spectrometers. Furthermore, OMI is described briefly, but not used in this study. On the other hand, different TROPOMI UVAI products are described and compared, but not which versions. The authors claim that two wavelength pairs are available. Since version 2 of TROPOMI L2 products, the ESA TROPOMI UVAI products also has a 335/367 nm pair available, outside ozone absorption, in preparation of the Sentinel-5 mission. Only in the version 1 data two UVAI wavelength pairs were used, which raises the question what version the authors have used and whether they are consistent. Version 1 was based on collection 1 L1b data, while version 2 is based on L1b data that was degradation-corrected. It is unclear whether the NASA and ESA TROPOMI UVAI products that are compared are even based on the same L1b input.

The AOD description is also elaborate, citing quite extensively from papers of the last two to three decades with seemingly random details. If anything, the section gives a good impression of the despair and indecision a naive user feels when seeing the sprawl of AOD products. Instead of making clear choices from the different AOD products that NASA and NOAA have developed, the products are put together in files, leaving it up to the user to decide what is best to use. The co-authors of this paper are the algorithm developers of the products and the absolute world-leading experts in their field, but even after this study the only conclusion is that AOD products are mainly inconsistent.

Having said that, the analysis of the performance of the satellite products over the Aralkum desert is decent. One of the interesting findings is the use of the 95 percentile of the UVAI to indicate a plume. The percentile is much less dependent on the wavelength choice or calibration or UVAI definition than a threshold method, which is an interesting find. Unfortunately, the authors do not give any recent references to the use of the UVAI as an indicator of plumes, even though the UVAI has been quite successful for that (e.g. Khaykin et al, 2022, Nature comm.). The ever recurring discussion is what threshold should be used, while using a percentile could be more consistent, which is unfortunately not further explored. Even more disappointing is the fact that the authors revert to a threshold method in their own analysis!

My main criticism is that no effort has been done to try and improve the AOD retrievals with the knowledge of the dust emissions. It is known that AOD retrievals suffer from (inaccurate) knowledge of aerosol type. Here, an extensive knowledge of dust type is claimed, judging form Fig 1., but none of this information is used. One would expect at least some tests using a more proper aerosol model to quantify the improvement that can be expected, or a choice of AOD product that is most suitable. Instead, the authors conclude "HAD the dust optical model be forcibly used in the retrieval, the EPS AOD at 0.55 µm WOULD have been significantly higher and brought into closer agreement with the DB and MAIAC algorithms." which is too general a statement and not new, to say the least. Who else should perform such an analysis then the algorithm developers? With missions like PACE and EarthCARE it can be expected to have better aerosol microphysical properties available in the near future, and it would be interesting to know what can be expected for the AOD products described here, and what is needed to improve them.

I conclude that the paper may be interesting for readers with 'a knowledge gap in the performance and consistency of (NASA/NOAA) satellite observations in characterising the saline dust emission from the Aralkum Desert', but I am not impressed.

Specific comments:

line 58: . Data users may struggle with the product choice, not knowing the strengths and limitations of different products when applied to their region of interest. 

Unfortunately, no recommendations are provided for users after this study. Nor are QA values updated or created as a result from this study. One result is that the AOCH is reliable for AOD> 1.8, which is a very high number. It is unlikely that users would benefit from such a threshold.

line 99: Historically, two wavelength pairs have been used: 340/380 nm and 354/388  nm.

Historically, more different wavelengths pairs have been defined, depending (mainly) on the instrument capabilities. With the introduction of hyperspectral instruments including the UV, the wavelengths became an actual choice, so the ozone-free wavelengths 354/388 nm were mostly chosen, next to the most used 340-380 nm. I suggest to add this to the review.

line 146:

ESA TROPOMI UAI is in three wavelength pairs. Which version has been used? version 2 UVAI is about 0.5 higher than version 1 data for the displayed area, due to degradation correction. What version of data is used for the NASA UVAI?

2.1.2 OMI UVAI is not used, why is it described? Either describe all available products, or the ones used in this study.

158: AOD provides aerosol burden in the atmospheric column ->

"AOD is the aerosol light extinction of the total atmospheric column" A proper physical definition is not out of place in an atmospheric physics paper.

line 212: strange sentence, please rephrase in a way that does not suggest a self-replicating algorithm.

line 248: Once dust is detected -> If dust is detected

line 324 Six EPIC bands are considered, including the oxygen A and B bands at 764 and 688 nm and two reference continuum bands at 780 and 680 nm. These are four (described) bands.What are the other two and why are they left out?

line 346: "they are subject to larger biases and usually not recommended for use in scientific studies. "

Wow! If scientific user shouldn't use them, who should? Non-scientific users?

line 398 UAVI = UVAI

line 398-400: "We create the co-located data by first identifying the nearest pixels from all UVAI products to the CALIOP footprints, and then shifting the UVAI pixels along track based on the sensor scan time differences.

I don't understand this. Please, elaborate.

line 455: The quality of retrieved AODs for heavy aerosol events tends to have large residuals and high possibility of cloud contamination.

I don't understand this sentence, please, rephrase.

Caption fig 6: MAIC = MAIAC

Caption Fig 6: The dust plume is indicated by the TROPOMI TropOMAER UVAI value of 2 (red contours). Why not the 95th percentile? I thought this worked better and was more consistent.

line 507. If "the non-negligible water-leaving radiance from these pixels may be erroneously treated as aerosol signal" in the VIIRS DT algorithm, why is MODIS DT not affected?

line 517: retrieves -> retrieve

line 550: Using the CALIOP 532 nm extinction-weighted height as ground truth"

Caliop is space-based.

line 557: degraded -> Reduced

lines 569-570: The extent to which the dust optical models deviate from the saline dust from Aralkum (e.g., refractive index, size distribution) has yet to be determined.

Why has this not been addressed in this paper?

Line 589-590: Why was the EPS algorithm not tested with the correct optical (dust) model? Would this improve the situation? Who else should test this? What is the use of a focus on a desert dust case with measurements on the dust properties and not using those measurments?

line 610: "This study highlights the need for in situ measurements of the physiochemical and mineralogical properties of the saline dust from Aralkum, which are critical for improving the representation of the regional aerosol optical models used in satellite aerosol algorithms.

We also call for routine ground-based aerosol measurements in the downwind region..."

The measurements available now are not used and the paper doesn't even show the possible improvement using a better dust representation. I suggest to start with that.

This paper documents analysis of aerosol properties over the newly formed Aralkum desert using multi-sensor satellite observations. Such study is indeed important as aerosol properties from this desert contain significant amount of salt as well as dust, which is unique and maybe be well represented by satellite retrieval algorithms. However, I read through the manuscript but was disappointed because nowhere in the paper discussed the difference between aerosol properties here and other desert regions. To me, this is the most important point of such a study. So I have only one major comment, which is, compare the optical properties of saline dust storms with regular dust storms. This can be done through:

1. It looks to me that the saline dust is more scattering than regular dust. Could the authors compare AAI or AAOD products from TROPOMI to verify this?
2. Is there any particle size difference between saline dust and regular dust? Perhaps take a look at MODIS AE, although it is not accurate over land, some qualitative clue may appear;
3. Do the retrievals of MODIS and VIIRS perform differently over saline desert and other deserts? All the AOD products used in the paper rely on assumptions of aerosol models. The authors did suspect that the uncertainty in these AOD products maybe associated with incorrect model assumptions, but more could be done to verify this. For example, how is the AOD performance here compared to regular deserts, such as North Africa and the Middle East? Could the difference be explained by the different aerosol absorption, vertical distribution or surface reflectance? Some retrieval experiments can even be done, if not too much difficulty, by adjusting the aerosol absorption at least.

In short, the authors should provide a comparison between the characteristics and retrieval performance of Aralkum desert and other regular deserts to gain insight into the unique characteristic of saline dust, as well as its satellite retrieval.

Finally, I wonder why the CALIPSO AOCHs from 532 and 1064 nm channels are different?

This paper presents a satellite dust retrieval intercomparison study over Central Asia, with respect to a series of dust storms produced by the Aralkum Desert in May 2018. The authors analyze data from multiple satellite instruments (OMPS, TROPOMI, EPIC, MODIS, VIIRS, CALIOP) and retrieval algorithms over both land and sea so as to understand the performance of the retrievals and the locations where particular caution is required in interpreting the retrieval output. In particular biases are apparent over the turbid waters of the Caspian Sea. Overall I find this paper to be acceptable and straightforward, providing a modestly informative intercomparison of UV and visible satellite dust retrieval products over the Central Asian region.

I think the biggest gap in this paper concerns the apparently saline nature of the dust storms being analyzed. The word saline appears very prominently in the title of the paper, so it is a bit disappointing that this forms such a minor aspect in the analysis and discussion. I see that this is discussed in a couple of paragraphs in the Conclusions, but actually the analysis in the paper is about the performance of the dust retrievals more broadly. The dust is assumed to be composed of a more general desert dust type. It therefore seems slightly misleading to state that this paper is about saline dust.

Moreover, I find it disappointing how few of the retrievals analyzed are over the Aralkum itself. The Aralkum has a relatively dynamic surface environment, with the area of the remaining lake varying both interannually and seasonally, and so I would expect this to be a challenging area for satellite dust retrievals that would be worth a more focused analysis. A related point is made by the authors themselves in the Conclusions around line 580 with respect to UVAI retrievals over the Caspian Sea. However most of the analysis of dust activity is performed over Central Asia more generally, and over the Caspian. This seems like a missed opportunity. Given that the paper is prominently declared to be about saline dust from the Aralkum, I feel that this subject would warrant a deeper analysis of retrievals in the direct vicinity of the Aralkum itself, such an analysis could be an extra subsection or even a whole section within the paper.

Specific comments

Line 1: “saline dust emission from…”. I don’t think it is necessarily the emission that is being characterized, rather it is the aerosol above the Aralkum. Perhaps fairer to say “saline dust aerosol emitted from…”

Line 51: I would say that dust activity over the Sahara Desert has also been frequently analyzed over the last few decades.

Line 350: it is worth mentioning here that the SEVIRI Dust RGB composite image is in the infrared, an important distinction with respect to the rest of the (UV and visible) data sources used in this paper.

Figure 3(g): I think it would be worth adding a vertical line (maybe dashed) to mark the UVAI threshold of 2 that is discussed in the paragraph starting on line 386.

Line 429: surely the wide geographical range (34-50°N, 47-70°E) refers to the Central Asian region, not the Aral basin.

Figure 5: the map panels are a little small here. I recommend transposing the panels into a more vertical configuration (i.e. 5 rows x 3 columns), this would make the maps bigger.