the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 28 Apr 2025
Derivation and validation of a refined dust product from Aeolus (L2A+)
Abstract. The missing cross-channel of the lidar system aboard Aeolus (Atmospheric Laser Doppler Instrument; ALADIN) prevents the generation of accurate optical products when depolarizing atmospheric layers are probed. The absence of the cross-polar component also limits ALADIN's ability to distinguish between different aerosol and cloud types, in its retrievals. To address these limitations, an enhanced Aeolus aerosol product, with a focus on dust, has been developed in the present study to support aerosol data assimilation in dust transport models and improve Numerical Weather Prediction (NWP). The enhanced aerosol product is derived through a series of intermediate processing steps that integrate spaceborne retrievals/products from multiple sensors, reanalysis numerical outputs, and reference ground-based measurements. Both the primary (L2A), and enhanced (L2A+) Aeolus optical products, in terms of profiles of backscatter coefficient at 355 nm are retrieved using four different algorithms, the Standard Correct Algorithm (SCA), the Standard Correct Algorithm at the middle-bin vertical scale (SCA-MB), the Maximum Likelihood Estimation (MLE), and AEL-PRO. These products are validated against ground-based reference observations obtained from the eVe and PollyXT lidar systems, operated as part of the ASKOS/JATAC experimental campaign in Mindelo, Cabo Verde. The approach is detailed on the basis of an indicative Aeolus overpass in the proximity of Mindelo on September 3, 2021, discussing ALADIN’s sources of underestimation in terms of L2A backscatter coefficient at 355 nm profiles in the presence of desert dust particles across all four retrieval algorithms and the induced improvements achieved by accounting for the missing cross-polar component. A statistical evaluation of all Aeolus overpasses during the entire ASKOS/JATAC campaign in the Cabo Verde/Mindelo region confirms the enhanced performance of the upgraded L2A+ product compared to the original L2A product. This improvement is evident in both Aeolus-eVe and Aeolus-PollyXT comparisons across all retrieval algorithms and is marked by higher regression slopes and lower bias scores. Specifically, among the algorithms, AEL-PRO and MLE L2A+ products show significant improvements in alignment with eVe lidar observations, with bias reductions from -0.46 to -0.17 Mm⁻¹sr⁻¹ (MLE) and -0.43 to -0.04 Mm⁻¹sr⁻¹ (AEL-PRO). They also achieve lower RMSE values (0.87 Mm⁻¹sr⁻¹ for MLE and 0.62 Mm⁻¹sr⁻¹ for AEL-PRO) and better regression slopes, increasing from 0.39 to 0.65 (MLE) and 0.53 to 0.87 (AEL-PRO). Similarly, L2A+ adjustments reduce biases and improve regression slopes in Aeolus-PollyXT comparisons, especially for SCA-MB and MLE algorithms. These advancements establish the enhanced L2A+ dust product as a strong candidate for aerosol data assimilation, supporting improved dust transport modeling and further enhancing Numerical Weather Prediction (NWP).
Competing interests: Some authors are members of the editorial board of AMT.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(3266 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-1175', Anonymous Referee #1, 19 May 2025
-
AC1: 'Reply on RC1', Konstantinos Rizos, 27 Nov 2025
1. We thank the reviewer for the constructive comments on the clarity and presentation of the abstract. All the points raised have been carefully considered and incorporated into a revised version of the abstract.
•Regarding the description of how the enhanced dust product is refined, the text has been rewritten to more clearly convey the methodology. “The L2A+ product is generated through a series of processing steps that integrate multi-sensor satellite retrievals for cloud screening and aerosol layer characterization, CAMS reanalysis outputs to classify aerosol types, distinguish dust from non-dust fractions, and provide the missing depolarization ratio values required for the Polarization Lidar
Photometer Networking (POLIPHON) technique, along with ground-based lidar measurements used for performance assessment.” This revision directly addresses the reviewer’s concern about clarifying how the dust product is improved.
•The sentence detailing the retrieval algorithms has been rephrased for clarity and conciseness. It now reads: “Both the primary (L2A) and enhanced (L2A+) Aeolus pure-dust backscatter coefficient profiles at 355 nm are retrieved using four different algorithms: the Standard Correct Algorithm (SCA), the Standard Correct Algorithm with middle-bin vertical scaling (SCA-MB), the Maximum Likelihood Estimation (MLE), and AEL-PRO.”
•To avoid any misunderstanding, we have clarified in the beginning of the abstract that the absence of cross-polarization measurements in ALADIN is that it was originally designed as a wind lidar without depolarization measurement capability. Therefore, it detects only the co-polar component of the backscattered signal, limiting the accuracy of optical products in the presence of depolarizing atmospheric layers.
We believe that the above-mentioned revisions address the reviewer’s concerns and improve the clarity, accuracy, and readability of the abstract.2. We thank the reviewer for this valuable comment and for highlighting the need for greater clarity and dimensional consistency in Equation 1. The section has been revised accordingly. The variable Cₚₘ₁₀ is now explicitly defined as the total aerosol mass concentration (μg/m³), while Xₚ denotes the mass mixing ratio (kg/kg) of aerosol type p (SS₁–₃ for sea salt, DD₁–₃ for dust, OM for organic matter, BC for black carbon, and SU for sulfate). The term inside the parentheses, (pₘ / (Rspec × T)), has been clarified to represent the air density (kg/m³), derived from the ideal gas law. Multiplying the aerosol mass mixing ratio by this term yields the mass concentration in kg/m³, which is then converted to μg/m³. This correction ensures full dimensional consistency and provides a physically sound description of the conversion process. The notation has also been standardized using Xₚ to denote the mixing ratio of each aerosol type, in accordance with the reviewer’s suggestion.
3. We thank the reviewer for this constructive and detailed comment. In the revised version of the manuscript, Section 3.3 (“Quantifying the Aeolus pure-dust component through reconstruction of its missing cross-polar term”) has been substantially rewritten to clarify the aerosol typing methodology and to replace the previous threshold-based approach.
In the updated framework, Aeolus dust identification is no longer based on fixed concentration or fraction thresholds (e.g., 1.3 μg/m³ or 50%), but instead relies on a physically consistent, POLIPHON-based methodology. The one-step POLIPHON technique (Pappalardo et al., 2014) allows the decomposition of the total aerosol load into pure-dust and non-dust components by exploiting particle depolarization ratios. Since Aeolus ALADIN does not provide depolarization measurements, CAMS reanalysis products were integrated into the L2A+ workflow to (i) classify aerosols along the Aeolus overpasses, (ii) distinguish dust from non-dust fractions, and (iii) supply the missing depolarization ratio values needed to apply the POLIPHON method.
A detailed spatiotemporal collocation between Aeolus and CAMS was performed using nearest-neighbor matching, with CAMS outputs selected within ±3 hours of each Aeolus observation. For each Aeolus BRC bin, CAMS aerosol mass concentration profiles were averaged over the corresponding vertical layer. These CAMS-derived parameters, dust mass concentration and dust fraction, were used to qualitatively identify the atmospheric layers dominated by dust. Over these layers, the missing Aeolus cross-polar backscatter component was reconstructed using Eq. (4): 𝛽totalcross=𝛿𝑐𝑖𝑟𝑐355×𝛽totalco
where 𝛽totalcross is the missing cross-polar contribution from dust and non-dust aerosols, and 𝛽totalcois the directly observed Aeolus co-polar backscatter. The particle depolarization ratio 𝛿𝑐𝑖𝑟𝑐355is derived from CAMS-based linear depolarization ratios through the POLIPHON relation (Eqs. 2–3), incorporating theoretical values of δ₃₅₅,d = 0.244 and δ₃₅₅,nd = 0.03 (converted to circular values 0.65 and 0.06, respectively).
This new formulation allows the reconstruction of Aeolus’s missing cross-polar term, enabling the retrieval of complete (L2A+) pure-dust and total backscatter coefficients. It also ensures that the aerosol typing is based on physically meaningful parameters, namely, particle depolarization ratio and backscatter properties, rather than on empirically defined thresholds.
Finally, to clarify the reviewer’s concern, the “median value of the entire dust mass concentration” refered explicitly to the median over all Aeolus BRC bins within the identified dust layers, providing a consistent measure of dust intensity across the profile.4. In the revised version of the manuscript, Section 3.3 (“Quantifying the Aeolus pure-dust component through reconstruction of its missing cross-polar term”) has been expanded to clarify this point.
To produce the L2A+ dust product, the reconstruction of the missing Aeolus cross-polar component is based primarily on CAMS-derived depolarization ratio values, rather than directly on the dust fraction. Specifically, CAMS data are used to estimate the particle depolarization ratio through the one-step POLIPHON methodology, which enables the separation of dust and non-dust components in a physically consistent way. This approach allows the derived depolarization ratio to implicitly reflect the degree of dust dominance within each layer, without applying an explicit dust-fraction weighting. Therefore, even in cases where the CAMS dust fraction is below 100 %, the retrieved depolarization ratio, and consequently the reconstructed Aeolus cross-polar term, accounts for the partial presence of non-dust aerosols within the mixture.
Regarding the reviewer’s suggestion on using lidar ratios, we fully agree that this parameter provides valuable information for aerosol typing. Although the present study focuses on reconstructing the Aeolus cross-polar signal and evaluating the improvement in backscatter retrievals, the extension of this framework to include lidar-ratio-based analysis and comparison with reference values from EARLINET datasets is foreseen as part of our future work.
Overall, the revised methodology ensures that mixed aerosol conditions are treated consistently through the CAMS-based depolarization ratio retrieval, while establishing a robust foundation for future inclusion of additional constraints such as the lidar ratio.5. We appreciate the reviewer’s suggestion. The figures have been revised using thinner, less bold lines for the uncertainty ranges, improving clarity and allowing individual curves to be more easily distinguished.
-
AC1: 'Reply on RC1', Konstantinos Rizos, 27 Nov 2025
-
RC2: 'Comment on egusphere-2025-1175', Dimitri Trapon, 30 Aug 2025
A Review of "Derivation and validation of a refined dust product from Aeolus (L2A+)" by Konstantinos Rizos et al., Atmos. Meas. Tech. Discuss., https://doi.org/10.5194/egusphere-2025-1175, 2025
The paper is well structured and provides valuable insights into the potential of Aeolus aerosol data product Level-2A labelled Baseline 16 for the characterization of dust particles. Kindly consider the comments attached which may help clarifying some points.
-
AC2: 'Reply on RC2', Konstantinos Rizos, 27 Nov 2025
1. We thank the reviewer for this comment. When averaging AEL-PRO measurements per BRC profile, a strict cloud-filtering procedure was applied. Specifically, any BRC containing at least one cloud-contaminated measurement, as identified by both AEL-FM and SEVIRI (MSG), was entirely excluded from the averaging. Furthermore, for the AEL-PRO algorithm, the retrieved cloud-free co-polar backscatter profiles were additionally filtered using the AEL-PRO classification product, retaining only BRC bins classified as tropospheric aerosols (index = 103). As a result, the averaged BRC profiles include exclusively cloud-free aerosol measurements, ensuring that localized clouds or contaminated features do not affect the derived statistics.
2. In our analysis, we applied a combined cloud-screening approach using both the Aeolus Feature Mask (AEL-FM) and the SEVIRI CLAAS-3 cloud-mask product to ensure more robust identification and removal of cloud-contaminated profiles. Although the L2A Baseline 16 processor includes a cloud mask at the BRC level, we chose this combined approach to benefit from the higher horizontal resolution (~3 km) of the AEL-FM product and the complementary cloud information from SEVIRI (~4 km, 15 min temporal resolution). The AEL-FM algorithm, included in the Baseline 16 processor used in this study, provides a feature detection probability index capable of distinguishing between clear-sky, optically thick clouds, and fully attenuated signals, allowing finer discrimination of small-scale cloud structures. The SEVIRI CLAAS-3 binary cloud mask, on the other hand, offers continuous and independent geostationary cloud observations, enabling cross-verification of cloudy scenes detected by Aeolus.
By combining these two datasets, we ensured a more conservative and reliable cloud screening, minimizing the risk of residual cloud contamination in the Aeolus optical profiles used for aerosol analysis.3. In the earlier version of the manuscript, the identification of dust layers relied on CAMS dust mass concentration (median threshold: 1.3 μg/m³) and dust fraction (50%) to classify Aeolus BRC bins as dust-dominated. However, these thresholds do not necessarily correspond to fully pure-dust layers, as the classified bins may still contain a mixture of dust and non-dust aerosols. This partially explains why the increase in the L2A+ total backscatter in Figure 10 did not appear as strong as the theoretical ~33% underestimation factor expected for purely depolarizing dust.
In the revised version of the study, we implemented an improved, physics-based aerosol typing scheme using the one-step POLIPHON methodology. Through this approach, CAMS reanalysis data were used to estimate the particle depolarization ratio, enabling the physical separation of dust and non-dust components in the aerosol mixture. This method inherently reflects the degree of dust dominance within each layer, without requiring an explicit dust-fraction weighting. As a result, even for partially mixed layers, the retrieved depolarization ratio, and therefore the reconstructed Aeolus cross-polar contribution, appropriately accounts for the presence of non-dust aerosols, leading to a more realistic enhancement of the total backscatter signal.
Additionally, following the reviewer’s suggestion, we have modified the color scale in Figure 11 (previously Figure 10) to a non-linear scheme, which enhances visual contrast and makes the increase in total backscatter after the L2A+ correction more clearly visible.4. In response to the reviewer’s comment, we would like to clarify that in the initial version of our methodology, the enhanced L2A+ dust product was derived by estimating the missing cross-polarized backscatter component using a fixed depolarization ratio value for dust as a constraint. For this reason, explicit tables or varying values of the circular and linear depolarization ratios at 355 nm were not included in the manuscript.
Specifically, we adopted a fixed particle linear depolarization ratio (δ₃₅₅,d = 0.244) for Saharan dust, which was then converted to its circular equivalent using Eq. (3). This value was taken from the DeLiAn database (Floutsi et al., 2022) and was considered appropriate for our study domain over the Atlantic Ocean, where Saharan dust is the dominant aerosol type.
However, as correctly noted by the reviewer, this fixed-value approach limits applicability to regions dominated by other aerosol types. Therefore, in the revised version of our methodology, we improved the L2A+ dust retrievals by incorporating CAMS reanalysis data. CAMS-derived aerosol information was used to estimate particle depolarization ratios dynamically through the one-step POLIPHON method, enabling a physically consistent separation of dust and non-dust components and the derivation of an improved Aeolus L2A+ dust product, analogous to the approach used in the CALIPSO LIVAS database. The new methodology is now ready to be implemented on a global scale, allowing the production of an enhanced Aeolus dust product with improved accuracy and consistency across different aerosol environments.
Regarding ground-based depolarization measurements, particle linear depolarization ratio data were available only from the eVe lidar system operated during the ASKOS campaign, while the PollyXT system did not provide depolarization measurements for the examined period. However, in our analysis, the depolarization ratio information used to reconstruct the missing Aeolus cross-polar component was derived entirely from CAMS reanalysis outputs through the one-step POLIPHON methodology, rather than from Aeolus observations themselves. For this reason, we did not perform a direct comparison between the CAMS-derived depolarization ratios and the eVe ground-based measurements, since the CAMS-based ratios represent modeled, reanalysis-driven estimates and not Aeolus-measured quantities. A comparison with eVe depolarization data would therefore not serve as a direct validation of the Aeolus-based retrievals but rather of the CAMS product, which lies beyond the scope of this study. This type of comparison could be incorporated in a future study focused on validating CAMS-derived depolarization ratios against both satellite-based and ground-based lidar measurements, further supporting the reliability of CAMS data for aerosol typing applications.5. We appreciate the reviewer’s observation regarding the Quality Check (QC) flags included in the Aeolus L2A Baseline 16 products. In the current study, the Aeolus QC flags were not applied to the SCA, SCA-MB, or MLE retrievals. This decision was made to avoid a substantial reduction in the number of valid bins per profile per BRC, which would have significantly limited the statistical robustness of the Aeolus–ground-based comparisons. As suggested, we have now clarified this point in the methodology section of the revised manuscript, explicitly noting that the QC flags were not used and explaining the rationale behind this decision.
6. Indeed, the recent implementation of the MLEsub product in the Aeolus L2A Baseline 16 (processor version 3.16) provides improved horizontal resolution by reducing the averaging scale to sub-BRC level, corresponding to approximately 15–18 km per sub-profile. As demonstrated by Trapon et al. (2025), this finer resolution significantly enhances the representation of aerosol layers and mitigates the impact of averaging dilution, particularly in regions with strong spatial variability or partial cloud contamination.
Although our present study relies on SCA, SCA-MB, AEL_PRO and MLE datasets at the standard BRC resolution, we recognize that the use of MLEsub could further improve the cloud-screening accuracy and aerosol retrieval consistency. Following the reviewer’s suggestion, we have added a statement in the revised manuscript version noting this open point and highlighting that future work will include a comparative analysis of Aeolus optical products at BRC scale (e.g., SCA, MLE) versus MLEsub, in order to assess the impact of horizontal averaging on the retrieval of dust and mixed aerosol layers.7. Specific comments: All these specific points have now been corrected as suggested by the reviewer.
-
AC2: 'Reply on RC2', Konstantinos Rizos, 27 Nov 2025
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 953 | 135 | 25 | 1,113 | 40 | 61 |
- HTML: 953
- PDF: 135
- XML: 25
- Total: 1,113
- BibTeX: 40
- EndNote: 61
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review comments for “Derivation and validation of a refined dust product from Aeolus (L2A+)” by Rizos et al.
The Aeolus lidar provides only the co-polarized component of the return signal, with the cross-polarized component unavailable. This limitation presents significant challenges in retrieving backscatter from highly polarized returns produced by nonspherical particles such as dust or cirrus clouds. To address this, the authors have developed an enhanced Aeolus aerosol product (L1A+), with an emphasis on dust backscatter coefficients. The algorithm used to generate the L1A+ product incorporates cloud screening based on the CLAAS-3 dataset, derived from observations by the Spinning Enhanced Visible and Infrared Imager (SEVIRI) onboard the Meteosat Second Generation (MSG) satellites. Aerosol subtyping for dust identification relies on the Copernicus Atmosphere Monitoring Service (CAMS) reanalysis dataset. To compensate for the absence of cross-polarized signals, the algorithm assumes a fixed depolarization ratio for pure dust, as defined in the study, to constrain the retrieval.
The L1A+ dust backscatter data were compared with ground-based lidar measurements collected during field campaigns in Mindelo, Cabo Verde. The enhanced L1A+ product showed improved agreement with the ground-based lidar observations compared to the standard Aeolus backscatter product. This improvement is valuable for more accurate aerosol data assimilation in dust transport models and for enhancing Numerical Weather Prediction (NWP) capabilities. I suggest the paper is published after revisions. My comments are given in below.
1. Abstract: It could be improved to convey the key points more clearly and concisely. By reading the abstract (and the conclusions), it is unclear how the dust product is refined and improved. Lines 26–28 state: “The enhanced aerosol product is derived through a series of intermediate processing steps that integrate spaceborne retrievals/products from multiple sensors, reanalysis numerical outputs, and reference ground-based measurements.” This could be rewritten more directly, for example, as: “The enhanced dust aerosol product is obtained by estimating the missing cross-polarized component using a fixed depolarization ratio for dust as a constraint. Collocated satellite measurements assist with cloud screening and aerosol subtyping.”
Another example, lines 28-31 state “Both the primary (L2A), and enhanced (L2A+) Aeolus optical products, in terms of profiles of backscatter coefficient at 355 nm are retrieved using four different algorithms, the Standard Correct Algorithm (SCA), the Standard Correct Algorithm at the middle-bin vertical scale (SCA-MB), the Maximum Likelihood Estimation (MLE), and AEL-PRO.”, which could be rewritten as: “Both the primary (L2A) and enhanced (L2A+) Aeolus backscatter coefficient profiles at 355 nm are retrieved using four different algorithms: the Standard Correct Algorithm (SCA), the Standard Correct Algorithm with middle-bin vertical scaling (SCA-MB), the Maximum Likelihood Estimation (MLE), and AEL-PRO.”
The abstract begins with “The missing cross-channel…”, which may give the impression that the absence of cross-polarization measurements in the Aeolus lidar resulted from a malfunction. It would be helpful to clarify—either in the abstract or in the introduction—that the cross-polarization channel was intentionally omitted by design due to the coaxial configuration of the Aeolus lidar, which was optimized primarily for Doppler wind measurements.
2. Equation 1: The term CPM10 should be clearly defined. Below Eq. 1, the text states: “In the above formula (Eq. 1), the sum of … multiplied by the dry air concentration inside the parenthesis (𝑝𝑚/𝑅𝑠𝑝𝑒𝑐 ∗ 𝑇) calculates the mass concentration …” This suggests that CPM10 represents the mass concentration in μg/m³, and the term in parentheses is the dry air concentration. However, it is unclear what exactly is meant by “dry air concentration” here. Based on the units provided in the paper, the expression inside the parentheses appears to have units of kg2/(J·m³), which is inconsistent. Please double-check the dimensional consistency of this term and clearly define what is meant by “dry air concentration.”
Additionally, the terms SS, DD, OM, etc., seem to refer to different aerosol types. However, it is unclear what specific parameter of each aerosol type is being represented (e.g., mixing ratio, mass fraction, volume concentration). To improve clarity, consider using a single character (e.g., f for fraction, c for concentration or X for mixing ratio) to denote the parameter and use the aerosol type as a subscript (e.g., XSS, XOM, etc). This would make the equation structure more interpretable and scientifically consistent.
3. Section 3.3 Aerosol typing method: It appears that the Aeolus dust layer is identified based on CAMS reanalysis products, with predefined thresholds of 1.3 μg/m³ for dust mass concentration and 50% for dust fraction. It seems that the dust fraction is derived using Eq. (1); if so, please clearly explain how this equation is applied in the analysis. Regarding the dust concentration, it is unclear whether it is based solely on CAMS reanalysis data, Aeolus measurements, or a combination of both. Please clarify the data source and methodology used.
Additionally, in Line 491, the phrase “the median value of the entire dust mass concentration” is ambiguous. What does “entire” refer to in this context? Does it mean all BRC bins, the full vertical profile, or something else? Please specify for clarity.
4. Section 3.4 Adjustment of the missing Aeolus cross-polar backscatter component: Equations (2) and (3) are valid for pure dust. However, since a layer is classified as dust when the CAMS dust fraction exceeds 50%, it may not consist entirely of dust. Could the dust fraction be incorporated into the retrieval to account for the presence of other aerosol types?
It would also be valuable to present lidar ratios derived from the L1A+ backscatter coefficient and HSRL extinction coefficient, allowing comparison with previously reported values.
A related question—perhaps beyond the scope of this paper—is whether the lidar ratio could also be used as an additional constraint for identifying pure dust, alongside the depolarization ratio?
5. Figures: The quality of the figures should be improved. Figures 12 and 13, in particular, are overly busy, making it difficult to distinguish individual curves. As a result, it is challenging to assess the agreement between profiles by visual inspection alone. To improve clarity, consider using error bars to represent uncertainties instead of plotting two separate error curves for each backscatter profile.