the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 25 Sep 2025
Characterization of the annual cycle of atmospheric aerosol over Mindelo, Cabo Verde, by means of continuous multiwavelength lidar observations
Abstract. This paper presents an analysis of the annual cycle of aerosol optical, and geometrical properties based on multiwavelength-Raman-polarization lidar measurements for Mindelo, Cabo Verde, from July 2021 to August 2023. A quality-assured data set of more than 70 automatically-calibrated lidar profiles was manually evaluated. For the first time, a two-year time series of, e.g., layer-resolved aerosol optical depth (AOD), lidar ratio profiles, and particle depolarization profiles are presented for Cabo Verde to characterize the complete annual cycle of aerosol in the planetary boundary layer (PBL) and in the lofted aerosol layers. The aerosol conditions over Mindelo are complex with different mixing states of dust and non-dust components. A strong annual cycle was found in the overall aerosol layer top height and the geometrical extent, the AOD, and the dust fraction of the lofted layers. Furthermore, the data was used to explicitly define aerosol-related seasons. The dust season, characterized by geometrically and optically thick lofted layers dominated by Saharan dust above a slightly polluted marine PBL, ranges from June to September. Aerosol occurs up to 7 km height. The seasonal mean lidar ratios and particle depolarization ratios at 355, 532, and 1064 nm are 32–34 sr and 0.02–0.05 for the PBL and 39–48 sr and 0.16–0.22 for the lofted layers. The mixing season covers the months November to March and is characterized by a large variability of aerosol, including mixtures of dust and smoke. The mean lidar ratios and depolarization ratios are 33–38 sr and 0.03–0.06 and 48–60 sr and 0.09–0.16.
- Preprint
(5146 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 02 Dec 2025)
- RC1: 'Comment on egusphere-2025-3344', Anonymous Referee #1, 27 Oct 2025 reply
-
RC2: 'Comment on egusphere-2025-3344', Anonymous Referee #3, 22 Nov 2025
reply
Review of the manuscript of Gebaur H. et al. (2025): “Characterization of the annual cycle of atmospheric aerosol over Mindelo, Cabo Verde, by means of continuous multiwavelength lidar observations”
General comments
The present study used two years of vertical profiles of aerosol optical properties derived from multiwavelength Raman-polarization lidar measurements to characterize annual cycle of aerosol over Mindelo city, in Cabo Verde Islands. The study focused on the characterization of aerosols optical properties in both PBL and lofted layers providing important results related to the seasonal dynamic of aerosols in the western portion of Tropical Atlantic. These long term lidar profile results over Cabo Verde represent a relevant contribution to the understanding of Saharan Air layer dynamic and provide a basis to explore comparison with long term measurements operated in Canarias islands and results from intensive experiments deployed in the past. No doubt, its content is within the scope of ACP. I recognize the scientific relevance of this study and importance of considering its publication on ACP after minor revisions and clarifications to the questions below.
The manuscript is well structured, it’s easy to read, although I often found that some parts of the writing need clarification and some plots could be improved (specific comments)
One major aim of the study was to define time frame of dust season, transition season and the mixing season over Cabo Verde the course of the two years. However, I’m a bit concerned regarding the frequency of one sample per week. Although the timeframe of this study is longer than previous intensive experiments, one sample per week in a two-year period seems to pose limitations to monthly characterization that support the seasons definitions. I would like to hear the authors on this. I wonder how representative the period is analyzed. I’m not sure that the authors said much about that. For instance, AERONET long-term monitoring could help with this.
Another aspect, attribution of aerosol scenarios related to non-dust sources, for instance smoke, could be improved with further support based on aspects such fire count across west Africa region during the study period combined with back trajectories of air mass etc.). I would like the authors to comment on that. This challenge to contextualize non-dust sources is present when the authors discuss the difference between this study results and previous experiments (those from SAMUM-2 and from Canarias.)
Specific comments
Ln 10: All types of aerosols or mainly dust?
Ln 11: correspondences between wavelengths and values need clarification.
Ln 11: What do you mean by mixing season, it is described in the text, but I think I need to be clear here in the abstract.
Ln 43: not “inner tropical” but Inter Tropical Convergence Zone.
Ln 65: suggestion to “columnar geometrical extension”
And I would add “dust fraction” as another item selected to characterize aerosol occurrence over Mindelo, which was critical in the discussion of aerosol annual cycle over Mindelo.
Ln 75: I think a map describing Mindelo location within Cabo Verde and in West Africa regional context would be very helpful for the readers.
Ln 78: I would say essentially the entire island is under marine influence given its small size.
Ln 108: “larger cloud-free period”, for how long typically?
Ln 127-128: Please, clarify about “measurements that are independent from each other”
Ln 167-169: How can this affect your analysis about the scenarios of distinct aerosol mixing? Maybe I missed it, I think I can see it somehow on the criteria that have been described, but I think that it could be clarifying to provide a conceptual definition of an aerosol layer that guides the study.
Ln 177-178: Let me see if I understood the average between the last AERONET instantaneous AOD from the day before and the first AERONET instantaneous AOD from the day after. Did you consider an average within certain interval to represent AERONET for the last estimate of previous day and the first of the day before? Instantaneous could fluctuate a lot. Maybe last and first hour mean.
Ln 182-183: why average instead of used spectral dependency, via Angstrom Exponent, between 440 and 675(or 870 nm)?
Ln 192-195: What are the difficulties of establishing a specific threshold to Mindelo conditions? Wouldn't background conditions help in that, or an alternative based on the gradient(slope) instead of a fixed value of backscattering coefficient?
Ln 207: Winter is a period of frequent transport events at lower levels in Cabo Verde, but for the period analyzed VDR seems not to show this feature. This led me to wonder, in terms of dust events over Mindelo/Cabo Verde, how close to (or far from) climatological behavior are these years analyzed? Maybe a comparison with AERONET climatology helps.
Ln 225-226: What would explain that? It would be interesting to hear about potential influence of smoke transport, since at this time there are fires around Sahel. And how about a possible effect of the threshold of the Hofer et al. algorithm during this period? Strong signal of clouds around 6 km during Jan to march, figure 1, can this play a role here?
Ln 255: Maybe a metric based on the ratio geometrical thick/ optical thickness could be interesting to explore and compare days aloft versus PBL.
Figure 3: the scale of date could be improved (rotate and increase resolution) to help the reader to make a clear connection between the specific dates mentioned in the text and the events discussed.
Figure 3: for AERONET suggestion to plot just the dots, without solid line, at least for those gap periods that there are no measurements. Plot (c) seems that it would be clearer with bar instead of dots.
Ln 260-264: Given that there are other aerosols affecting the dataset, it would be important to make clear at the beginning of the discussion. I mean, contextualize those profiles that are not exclusively dominated by dust at the beginning. Because I have this impression that the discussion to this stage focused on dust as it was the only aerosol type present. For example, that can be seen in the caption of figure 3c and 3d. It is dust most of the time, but it is not just dust. And there is also volcanic plume as cited.
Ln 270-273: As mentioned, it is difficult to follow this specific date discussion with the x-axis scale for date. You could rotate the axis 45 degrees and increase temporal resolution
Ln 275-277: Could you clarify here? Do you mean value above which you assume dust?
Ln 298: I would expect an increase in Angstrom exponent; I'm surprised that it did not happen. Any reason?
Figure 4, plot (a): I think it would be interesting to set a line based on the average value, it would help to track the variability by comparing dots with the average(line).
Figure 5: Somehow, I'm surprised that a clear dominant dust signal is absent in the PBL during winter. The persistence of dust events within PBL during winter period is a well-known seasonal feature in Cabo Verde region. Can the limited number of lidar profile sampling for this season play a role in this?
Ln 340-342: It seems that smoke has been important aspect to discuss Figure 5, but few were said about seasonality of biomass burning in West Africa and about fire and smoke leaving the continent during the sampling period.
Ln 345-346: How many measurements do you have per month? Considering that you have 1 profile per week, how can that affect this attribution?
Ln 347: Why 0.3 at 532 nm? Please, clarify this criterion.
Ln 356: what time of the year are you referring to?
Ln 356: Which cases are you referring to? Please, clarify.
Ln 362: Please, clarify: August, is it in the dust season range and classified as transition month?
Ln 375: I think to be more specific about the season that you are referring to would make reading easier. "Seasonal" term is used many times to refer to a specific period (dust season, mixing season), this can confuse the reader.
Ln 417: Considering the large sampling that this study collected did you think to evaluate diversity in dust intensive properties in the lofted layers related to dust sources origin?
Ln 419: What do you mean by Pollution of the dust?
Ln 472: What do you mean by African aerosol?
Citation: https://doi.org/10.5194/egusphere-2025-3344-RC2 -
RC3: 'Comment on egusphere-2025-3344', Anonymous Referee #4, 02 Dec 2025
reply
The authors present a first aerosol climatology over the measurement site of Mindelo, which is strategically located on the dust transportation pathway from Aftica to America. Dust-related seasons are defined based on the aerosol intensive properties in the boundary layer and within the lofted layers. The study provides useful results for communities related to aerosol typing/mixing, seasonal dynamics, and dust transportation. Important aerosol transport events such as the volcanic aerosol of La Palma can also be easily tracked from such a timeseries. The manuscript is well written and worth publishing. The figures are presented in a compact way that combines a lot of information without causing confusing. Below there are some specific and technical comments for the authors to address.
---------------------------------------------------------------------------------------------------------------------------------------------------------------------
Specific comments:
Line 98-99: “a minimized systematic relative error of 15 % for the particle backscatter coefficient”
Please shortly mention here where does this error come from (e.g. Ångström assumption, scattering ratio at reference height assumption?) and how was it assessed (e.g. analytically or with simulations?)
Line 100-101: “The errors of the lidar ratio and of the Ångström exponent were calculated via the Gaussian error propagation.”
Please specify wether this was an analytical or a Monte Carlo error propagation
Line 114-115: “Thus, the automatically-retrieved profiles of the aerosol optical properties are averaged over time periods between 15 min and 1 h, depending on the cloud conditions
It is not clear here if 15 min and 1 h averages can co-exists in an aerosol scene or whether one of the two sampling rate is used depending on the cloud conditions. Keep in mind that producing later on climatological values by 15 min and 1 h averages will lead to inconsistencies regarding the random error as 15 min average will always be more noisy. This will lead to larger error bars which is safe but not optimal. How is this handled in the climatology?
Line 145-150: Please provide here also the number available profiles per extinction and backscatter product. In addition, please specify if particle backscatter refers to pure Raman backscatter for 355 and 532 products (where any Klett profiles also included in the dataset?). Is the particle backscatter at 1064 nm solely produced by the Klett method?
Line 165: “The PBL top height was defined in the middle of the first significant gradient of the backscatter coefficient.”
Is this also done by manual inspection? Please specify.
Line 173-175: For the defined aerosol layers, layer mean values were calculated for the intensive properties (lidar ratio, Ångström exponents,particle linear depolarization ratio, and dust fraction), while the extensive optical properties (total and dust backscatter co-efficients and particle extinction coefficient) were integrated vertically.
Suggestion: An alternative way to produce the intensive averages in the layers is to produce them from the integrated extensive optical properties. This is more or less equivalent to producing an intensive property average by weighting with the corresponding extensive properties. The benefit compared to averaging the intensive properties directly is that the random error of the average is less affected by weaker parts of the profile that naturally have higher relative uncertainties.
Line 183-185: “For the uncertainty of the layer mean optical properties, the layer means of the errors described in Sect. 2.1 were calculated, while for the integrated values these errors were used as input for the Gaussian error propagation.”
This part is not very clear. A lot of data-related parameters were introduced in section 2.1. The authors should clarify to which errors they are referring to. In addition, the Gaussian error propagation method was mentioned previously without a specific explanation. Please provide some more information here. Is it based on Monte Carlo simulations or analytical formulas were used?
Lines 216 – 229: Are the averages (dots) weekly (Fri/Sat)? It will be helpful to remind the reader once again that each dot correspond to a Fri/Sat average.
Lines 227-229: This part could be written more clearly. The algorithm is simply sometimes confused due to lofted aerosol layers leading to higher uncertainty. In such cases the upper bar should be more trustworthy than the mean because of the presence of the lofted layers, which is the only source of such big error bars.
Figure 2 and lines 221-222: “as well as the temporal standard deviation of the seasonal mean layer top
heights”
According to the authors the uncertainties come from also from temporal averaging but this is not what is seen in figure 2. The error bars are mostly <100 m wide. However, from week-to-week the uppermost aerosol layer top can change be 2 km or sometimes even more. This contradicts the very small weekly variability. Is the temporal standard deviation really deployed?
Figure 3 and lines 252-266: There are some cases where the AOD retrieved by the lidar is higher than the sunphotometer AOD. Can the authors comment a bit on that? Is is due to diurnal cycle differences?
Lines 267-281: I would suggest that the Polyphon part goes first here so that the readers understand where the dust-related properties come from. In addition, the authors mentioned that a PDR of 0.31 was used. Was the 0.25-0.25 PDR range used for calculating the errobars of Figure 3c?
Lines 285-289: The seasonal pattern that the authors describe here it is not so easy to see, at least not for the lidar ratio and for the angstrom exponent that have high uncertainties. Indeed the PDR seems to show some difference between Nov and April for both 2021-2022 and for 2022-2023 in the lofted layer. But for the other 2 parameters I find it really hard to discern a repeating pattern. Indeed the LR difference at 355 and 532 is higher during 2022-2023 but for 2021-2022 this is not so clear. The authors could refer to the monthly averages that show more clear patterns
Line 307-309: This sentence is not so clear and needs to be refined. Which are these eight data points?
Line 359: The Angstrom exponent of the lidar ratio is not a very common lidar product. It should be introduced at some point in the paper before being used to define a threshold. The authors should also shortly describe what high/low values represent.
Line 395-428: In this part, the results between different past and modern campaigns are discussed. In order to really compare the values and draw firm conclusions one has to consider also the corresponding uncertainty. The authors should keep in mind that differences can only be significant if the uncertainty is sufficiently small.
---------------------------------------------------------------------------------------------------------------------------------------------------------------------
Technical comments:
Line 85: Typo here: “Aeolus was a equipped…”
Line 94-95: “one for far-range (fr) and one for near-range (nr) measurements”
I would suggest to define those 2 acronyms (fr and nr) with capital letters (FR and NR) to be more easily recognizable.
Line 102: “a newly calibrated instrument”
Please replace with: “a newly calibrated sunphotomer” (to avoid confusion with the lidar)
Line 112: “0.7 × 105 MHz m”
Should the units here be MHz m^-1 ?
Line 122-123: “The resulting smoothing length is 382.5 m for the nr-measurements and 742.5 m for the fr-measurements, which are the standard values used in the processing chain.”
Please provide here the 2 window sizes also in range bins.
Line 125: “The basis for this study were data” -- > “The basis for this study is data”
Line 142: “and any badly calibrated profiles” -- > “and any profiles with calibration issues”
Here, the sources of the calibration issues that the authors have experienced can also be mentioned in parenethesis (e.g. polarization calibration, Rayleigh calibration etc)
Lines 188-190: Are these results presented somewhere in more detail? Please point to the corresponding section
Lines 206: “On only few days,” -- > “Very rarely” (or similar expression)
General comment: I would suggest using an abbreviation for north hemispheric (e.g. NH) because it is repeated often and becomes distracting.
Line 256-257: “caused only small values of the AOD” -- > “contributed to only a small fraction of the AOD”
Line 270: “up to 0.5 was found” -- > “up to 0.5 were found”
Figure 3 and table 1 : Please rename “Dust Fraction” to “Dust Integrated Backscatter Fraction” (or similar)
Line 304: “the particle linear higher depolarization ratio” -- > “the higher particle linear depolarization ratio”
Citation: https://doi.org/10.5194/egusphere-2025-3344-RC3
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 950 | 89 | 23 | 1,062 | 16 | 16 |
- HTML: 950
- PDF: 89
- XML: 23
- Total: 1,062
- BibTeX: 16
- EndNote: 16
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review on the manuscript of Gebaur H. et al.: “Characterization of the annual cycle of atmospheric aerosol over Mindelo, Cabo Verde, by means of continuous multiwavelength lidar observations”
General comments
The manuscript presents the results on the analysis of the annual cycle of atmospheric aerosols over Mindelo, Cabo Verde using weekly PollyXT lidar observations from a 2-year dataset (July 2021 – August 2023) and collocated AERONET measurements. Moreover, the authors make use of the POLIPHON method for the derivation of the dust fraction and the DeLiAn database to support their analysis and findings. This study focuses on the investigation of the temporal development of the geometrical and optical properties among the different aerosol layers that are being identified (Planetary Boundary Layer and the lofted aerosol layers) for each measurement case aiming to define aerosol-related seasons such as the dust season, the mixing season, and the transition one. Even though I am a bit surprised about the low particle linear depolarization values at 355 nm (~0.16 ± 0.07) being reported during the dust season, I think that this study provides detailed information, that our community could benefit from, about the predominant aerosols and their geometrical and optical properties above Cabo Verde Islands, a region in the outflow of the Saharan desert.
Overall, the manuscript is well structured and well written but some parts could be further explained and/or discussed. The scientific significance makes the manuscript suitable for publication in AMT, after some minor revisions have been considered from the authors.
Specific comments
Line 13 “The mean lidar ratios … and 0.09-0.16”: I guess the authors provide the lidar ratio and depol. ratio values for the PBL and the lofted layers. Please clarify in the text.
Lines 40 – 43 “For example, SAMUM–2 … on the African continent.”: I would suggest to add an indicative citation for this statement.
Line 50 “these observations”: These observations refer only to Barreto et al. or SAMUM are included too? Please clarify.
Line 84 “ESA’s satellite Aeolus”: Here you could also cite some of the historical Aeolus mission papers (e.g. Stoffelen et al., 2005, https://doi.org/10.1175/BAMS-86-1-73; Reitebuch et al., 2012, https://doi.org/10.1007/978-3-642-30183-4_49). Moreover in the following sentences you could also add one indicative citation for the Aeolus wind products and one for the aerosol products.
Lines 94 – 95 “The PollyXT … (nr) measurements”: Please provide the full overlap height for the fr and the nr measurements.
Line 119 – 120 “Furthermore, … Tesche et al. (2009)”: Please provide the thresholds you used in the lidar ratio and the particle linear depolarization ratio for the dust and non-dust separation.
Line 145 – 150: It is not so clear many measurements have been used in the study after all. The authors mention a dataset of 73 cases (Fri/Sat dataset) which, however, to my understanding it is further reduced for the analysis presented from Figure 3 and on (i.e. sections 3.2, 3.3, and 4). More specifically, for this analysis, the authors use cases where both the nr and the fr backscatter and extinction coefficients (at 532 and 355 nm) should be available which may reduce the number of 73 cases. Please clarify and elaborate on possible change of the dataset being used in the analysis of each section.
Line 166 – 167 “lofted aerosol layers … Ångström exponent.”: Please elaborate on how the fr-particle backscatter coeff. is being used for the definition of the lofted layers. According to lines 167 – 169 only the depol. ratio and the BAE are used as metrics.
Line 170 “noticeable”: What does this mean? Please give more details or threshold values that indicate a noticeable difference.
Line 176 “… lowermost available value to be constant down to the ground.”: which altitude value is this for the backscatter coeff. and for the extinction coeff.? How do you define the height bin with the lowermost available value of beta and alpha (using full overlap height for nr)?
Lines 177 – 178 “The sum of the layer-AODs was compared with the columnar AERONET AOD.”: Please add which AERONET AODs are used for the comparison with the lidar-retrieved AODs from 532 and 355 nm.
Line 216 “The temporal development … illustrated in Fig. 2.”: which lidar profiles (and wavelength) have been used to create Figure 2 (PBL top height and lofted layer base/top heights from manual inspection vs from Hofer et al.)?
Lines 217 – 218 “Additionally, the manually-defined layer top heights are added in red for the
cases having no particle backscatter coefficient at 532 nm”: what about the availability of part. backscatter coef. at 355 nm? Why not to use also these profiles?
Lines 216 – 229: Why do the automatically retrieved layer top heights are being showed and discussed here, since the manual retrievals with visual inspection is considered as the best approach? Since the automatic method (Hofer et al., 2020) has its limitations for Mindelo cases (Lines 152 – 163, 192 – 197), please update figure 2 and corresponding analysis and discussion using the manually retrieved layer top heights. The current Figure 2 should be moved in an appendix and could be updated by showing both the automatic and manually retrieved layer top heights, if the authors aim for an investigation of the applicability of Hofer’s method to Mindelo.
Figures 2 – 5: A suggestion for better visualization would be to consider adding gridlines (at least for the major ticks) in the plots. Especially for the time series figures I think it would be helpful to add minor ticks in x-axis to indicate the non-labeled months (Aug, Sep, Nov, Dec, etc). Same applies for figures in the Appendix.
Line 286: Please revise the sentence to be more clear.
Line 292 “… slightly below 0,…”: please provide an average or indicative number.
Figure 5: Since the monthly mean values per year are not statistically significant, then why not using all data points?
Lines 316 – 318 “A cluster of data points… Floutsi et al., 2023)”: To my understanding this cluster is cluster 1. Please integrate the cluster numbers (cluster 1 – 6) throughout the whole paragraph to be easier to the reader to link Fig. 5 with the discussion (lines 313 – 342).
Lines 325 – 326 “According to Floutsi et al. (2023), data points with a particle linear depolarization ratio larger than 0.25 point to the occurrence of pure dust in the lofted layers”: Here the authors use the threshold value of 0.25 for defining the pure dust cluster (cluster 6) but this contradicts the use of particle linear depolarization of 0.31 for the pure dust separation in the POLIPHON products in section 3.2 which leads to the point of existence of a non-dust contribution in all lofted layers (lines 276 – 281).
Line 349 “mean particle linear depolarization ratio”: please provide the wavelength.
Lines 351 – 352 “Similarly… dust season”: I am confused. Is this an additional criterion or the result after applying the 3 criteria from lines 347 – 350 into the measurements for these months? If it is an extra criterion, does this mean that measurements from the months of the dust dominated cluster do not fulfill the 3 criteria?
Line 354 “transition months”: What are the criteria for defining a month as a transition one? Please include a short description in the text.
Lines 357 – 359: Are these conditions/criteria being used to define a mixing season? Different parameters are being used to define a dust (total AOD, contribution of lofted AOD, particle linear depol. ratio) and a mixing season (dust fraction, Ångström exponent of lidar ratio). Could the authors support why they do not use common parameters (with different threshold values) to distinguish the aerosol related seasons.
Line 359: The Ångström exponent of the lidar ratio is never mentioned before in the analysis or in the plots, but is used here as a criterion. Could the authors elaborate on why they do not use the backscatter or extinction related exponents?
Line 362 “August”: Do you mean October?
Line 368 “is 5.6±0.9 km (automatically-retrieved layer top heights).”: Again here, why do the authors use the automatic retrieved layer top heights instead of the manually retrieved ones which are considered as the best approach according to section 2.3.
Line 393 “using the two-layer approach according to Berjon et al. (2019)”: I would suggest to add a brief comment about the main differences/assumptions of the Berjon et al., compared to the direct retrieval of the lidar ratio from the lidar.
Line 413 “… and a lower particle linear depolarization ratio …”: I would like to see a discussion from the authors on the differences in the depol. ratio (mainly for 355 nm) with respect to the SAMUM-2b results (and the DeliAn database to my opinion). Currently only the lidar ratio differences are being discussed.
Line 454 “(depth of around 4 km and AOD up to 0.5 at 532 nm)”: It is not clear if these values correspond to seasonal mean values. Please clarify and use seasonal means (if not the case).
Lines 451 – 465: here the authors could also add key conclusions also for the rest intensive optical properties (e.g. lidar ratio, particle depol. ratio) for each aerosol-related season.
Technical corrections
Line 48 “multiple complete years”: I would suggest an addition like “multiple complete years above Canary Islands” or similar.
Line 80 “Aeolus was a equipped with a wind…”: Is something missing here?
Line 112 “0.7 x 105 MHz m.”: is the “m” at the end a typo?
Line 143 “In most cases, single optical properties at a certain wavelength and only one complete nighttime measurement”: Looks like something is missing here. Please revise.
Line 234 “The latter PBL top heights are often …”: Suggested change to " The highest observed PBL top heights (*maybe add here the range of PBL top heights for the cases with no lofted layers) are often…" or similar?