This manuscript presents a comprehensive and valuable update of the MAPIR algorithm (v5.1) for dust aerosol retrieval from IASI observations. The work is technically solid, the dataset is highly relevant for the community, and the effort to build a long-term, consistent TIR dust product is particularly commendable. The paper also demonstrates a strong level of expertise and provides a thorough description of the methodology and evaluation.

However, despite these clear strengths, I have several major concerns that should be addressed before publication. In my view, some of these issues are substantial and currently limit the robustness and interpretability of the results. In particular, questions related to the treatment of spectral noise, the handling of known issues in the IASI-C dataset, and the validation strategy need to be clarified and/or strengthened.

For these reasons, I recommend major revisions. I believe that addressing the comments below would significantly strengthen the manuscript and make it a strong and impactful contribution to the field.

Major comments:

• IASI-C - different pre-processing
• This appears to be a blocking issue. Publishing a dataset affected by such a bug raises concerns. I would strongly recommend either resolving this issue before publication or restricting the study to IASI-A.
• Length and readability
• The paper is very long and highly information-dense. Several sections could be streamlined to improve readability without losing scientific content. For example, L305–315 could be simplified, and L317–319 may not be necessary. Overall, reducing overly technical detail in some parts would make the manuscript more accessible.
• Treatment of spectral noise (L345–350)
• The artificial reduction of spectral noise does not increase the information content of the retrieval. The information content is governed by the Jacobians (i.e. sensitivity of the signal to the state variables), not by the noise itself. What is actually modified here is the balance between observation and prior in the cost function. While this may improve convergence, it is not physically justified. The spectral noise should reflect the true instrument characteristics, not be tuned to fit the retrieval. Please clarify how the values used in v5.1 were determined.
• Vertical profile validation
• It is stated that another manuscript is under preparation. I would then suggest publishing these 2 papers together, as companion papers. Section 3.6 (profiles) would be better placed in that companion paper rather than included here.
• Dust AOD Validation
• The validation effort is appreciated, especially given the known challenges of TIR aerosol validation. It should be stressed here that these results should indeed be considered qualitatively, as a proper comparison is not directly possible. However, it is clear that the a priori of 2 particles cm-3 leads indeed to an overestimation for low AOD. How do these results compare to the previous version of MAPIR and to other dust retrievals from IASI (e.g. DLR IMARS, LMD, ULB)? Given that this paper focuses on algorithm improvements, such benchmarking is essential.

Minor comments:

• Ensure all acronyms are consistently defined (e.g. OEM, AOD appear multiple times). Consider using a glossary package (e.g. glossaries in LaTeX).
• Add DOIs and clickable links for all references where possible.
• Figures are often too small and difficult to read:
• Increase font sizes (comparable to main text)
• Harmonise figure styles
• Use more descriptive, self-contained titles (e.g. include instrument names directly in titles)

Specific comments

• L26 remove extra punctuation → “wind speed (e.g. Choobari et al., 2014)”
• L41 best -> highest
• L113 Why mention the OEM here?
• L126 I would clarify early that the setup described is for the climate setup
• L126-132 Clarify the reasons behind the different setups. Also, what does it mean “when the OE doesn’t perform well”? How is this evaluated?
• L139 spherical assumption - please mention already here that this assumption is valid in the TIR but less in the VIS (and keep reference to section 3.2)
• L143 high-resolution, what does it mean exactly?
• L146 The resolution of MAPIR retrieval vertical grid should be specified here to ease readability
• L163 Does it make sense to use a fixed concentration for all other gases? How much gas absorption is there in the TIR? What are the default concentrations in RTTOV? What is the impact of assuming the surface Lambertian in the TIR?
• L173 The “negligible effect” should be demonstrated
• L174 snow fraction set to 0 -> Is it a valid assumption? In which cases?
• L210 no need to repeat the OEM acronym
• L226 - 232 this paragraph might be technical and not necessarily needed in an already quite long scientific paper
• L246 PSD not defined
• L255 how is this uncertainty reported then?
• L272: clarify → “not prefiltered in MAPIR v5.1”
• L278 “decent” quality is vague
• L282 is there some validation of this filter?
• L284 Provide a reference or an expanded explanation
• L289 was shown - where?
• L292 wrongly categorizing intense dust events as clouds - provide reference
• L296-298 The very small impact of reducing streams seems surprising → needs justification
• L305-315 unclear → needs clarification or simplification
• L335 -337 If not quantified, consider removing
• Figure 2 is too small, hard to read
• L362 - 368 Only ~35% of retrievals pass final QC → is processing cloudy scenes justified?
• L371 This shall be explained before the publication of this paper. At least, it would be nice to see that some investigation has been performed and possible factors excluded
• L372 Again, this is a blocking issue for publication
• Figure 3 is too small
• L533 If it really only slightly changes the statistics, why do you remove the worst 3%? Especially if this 3% includes high values of AOD, the statistics (both correlation and RMSE) might change significantly. I would refrain from such a filter, especially given that the data entering the comparison is already pre-filtered.
• Figure 7 day/night distinction unclear → use different markers
• Redundancy between Figures 7 and 8. Also, from Figure 7, it looks like there should be some colocation for IASI-A. Clarify why IASI-A colocations seem to be missing.
• L741 Could Jacobians be computed numerically (perturbation approach)? Shouldn’t sensitivity also depend on other parameters (e.g. gas concentrations, AOD magnitude)?
• L790: Could reduced IASI-C coverage be explained by erroneous cloud filtering?
• L793: If discussed later, avoid mentioning prematurely
• L794–797: Clarify reference to Figure 18

Review of “The Mineral Aerosol Profiling from Infrared Radiances version 5.1 algorithm and its evaluation”

This study describes version 5.1 of the MAPIR algorithm for retrieving dust concentration profiles and column-integrated AOD at 10 µm from IASI observations onboard the Metop-A and Metop-C satellites. Key updates from version 4.1 include extending the vertical retrieval range to 10 km, removing pre-filters to maximize data coverage, adopting updated surface emissivity and radiative transfer models, reducing spectral noise inflation, and propagating temperature and humidity uncertainties. The algorithm is evaluated against AERONET, CALIOP, and EARLINET. Long-term stability and IASI-A versus IASI-C consistency are also assessed. Overall, the paper is comprehensive, transparent about limitations, and delivers a scientifically valuable long-term data product. However, there are still some major converns needed to be addressed before considering acceptance.

1. Separation of retrieval quality from the TIR-to-VIS conversion uncertainty. In sect. 3.3, the authors compare MAPIR converted dust AOD at 550 nm against AERONET SDA coarse-mode AOD at 500 nm, using a fixed conversion factor of 1.78. The authors mentioned that the AOD-dependent bias could be a constant bias if a different conversion factor (2.85) were used. This means that the central validation metric conflates two distinct sources of uncertainty: the quality of the 10 µm retrieval itself and the accuracy of the wavelength conversion, which depends on assumed particle size and refractive index (Capelle et al., 2018; Clarisse et al., 2019; Zheng et al., 2023, 2026).
   
   
   
   As the authors acknowledge that direct reference to AOD at 10 μm is currently unavailable, it is suggested to provide a sensitivity analysis in which the comparison statistics (correlation, slope, bias, same as Figures 4 and 5) for a range of plausible conversion factors (e.g., 0.9 to 3.5, as quoted from Clarisse et al., 2019), showing how the statistics change. In addition, it is suggested to acknowledge the necessity of constraining dust microphysical properties (e.g., size, shape, refractive index) to reduce the uncertainty contributed by the conversion factor, as previous studies mentioned (Capelle et al., 2018; Clarisse et al., 2019; Zheng et al., 2023, 2026).
   
   
   
   It would also be informative to note which end of the conversion factor range corresponds to which particle sizes, such as "smaller particles produce higher conversion factors,” which would help the reader build intuition.
2. Quantitative profile validation needs strengthening. The profile comparisons with CALIOP (Sect. 3.6.1, Figs. 12–14) are presented as qualitative, side-by-side visualizations without summary statistics. The EARLINET validation (Sect. 3.6.2) is limited to 12 dust events at one station (Limassol, Cyprus). Given that MAPIR’s unique selling point relative to column-only TIR products is precisely the vertical profiling capability, a more rigorous profile validation would significantly strengthen the paper.
   
   
   
   Therefore, for the CALIOP comparisons, it is suggested to add layer-by-layer statistics (mean bias and correlation) for the matched orbit segments, even if the comparison is approximate due to different retrieved quantities (extinction vs. concentration). If possible, for the EARLINET comparisons, the authors could also consider expanding the analysis beyond Limassol to other EARLINET stations in the dust belt. Otherwise, it would be better to provide the mean and standard deviation of the MAPIR–lidar difference as a function of altitude for all 12 cases, with and without the averaging kernel correction, to provide a quantitative sense of the systematic vertical bias.
   
   
   
   Although the authors acknowledge that a dedicated profile validation paper is in preparation, a brief quantitative summary here would improve the current manuscript substantially.
3. The a priori positive bias floor and its implications for climatological averages. The increased minimum a priori concentration of 2 particles cm⁻³ across the entire profile (Sect. 2.1.3) creates a retrieval floor of approximately 0.06 at 10 µm AOD (~0.14 at 550 nm), which the authors acknowledge (lines 558–565 and 835–836). While this is understandable from a retrieval stability perspective, the consequences for users constructing climatologies, computing regional means, or performing trend analyses are significant. In dust-free regions, the product would always report non-zero dust AOD, inflating global or regional averages. I suggest the authors consider whether a bias correction or an additional quality flag (e.g., a flag indicating that the retrieval has not moved meaningfully from the a priori) could be included in future data versions.
4. The MAPIR retrieval uses a single set of dust optical properties: a log-normal size distribution with 0.6 µm mean radius, a refractive index from GEISA–HITRAN, and spherical particles (Sect. 2.1.2). However, previous studies have shown that there are significant global variations of dust size distribution (Formenti and Di Biagio, 2024) and dust TIR refractive index (Di Biagio et al., 2017). It is suggested to add discussions on whether the use of regionally varying refractive indices (as done by the Di Biagio et al. dust CRI database) could improve the retrieval, at least as a future development. If such sensitivity tests have already been performed for earlier MAPIR versions, referencing those results here would suffice.
   
   
   
   Although assuming dust to be spherical has been demonstrated to have similar optical properties in TIR as in non-spherical shapes, the authors also calculate the conversion ratio to visible based on the spherical assumption. The impact of dust non-sphericity is not negligible in visible (Huang et al., 2020; Saito et al., 2021). A discussion of whether the spherical assumption introduces systematic biases in the retrieval and in the conversion factor calculation is needed.

Minor comments:

Abstract, line 5: The abstract mentions retrieval of “10 µm AOD and mean altitude” but does not mention that the evaluation is performed using the 550 nm converted AOD. Since most readers will use the 550 nm product, briefly noting the conversion and its associated uncertainty in the abstract would set appropriate expectations.

Sect. 2.1.5, lines 170–178: The surface emissivity bias in CAMEL v2 over the Sahara during summer is noted but not quantified in terms of its impact on AOD retrieval. Since the surface emissivity uncertainty analysis in Sect. 3.7 shows up to 0.05 AOD uncertainty; the authors should discuss whether this uncertainty is dominated by the CAMEL bias or by intrinsic emissivity variability. If the former, would using an alternative emissivity product (e.g., MODIS-based monthly emissivity) reduce this uncertainty?

Sect. 2.2.4, lines 285–300: The reduction from 8 to 4 DOM streams is validated with a maximum AOD difference of 0.015. It would be helpful to state the conditions under which this test was performed (e.g., dust AOD range, viewing geometry). At large viewing angles or high optical depths, the sensitivity to stream number may increase.

Sect. 3.3.3, Figure 6: The colorbar range (0.1 to 1.0) is too smooth to tell the variation. Since most stations have R > 0.6, consider using a narrower range (e.g., 0.4 to 1.0) or a colormap with limited intervals to better distinguish good from mediocre stations.

Sect. 3.9, IASI-A vs. IASI-C: The discussion of the Metop-A orbital drift (lines 793–809) is important, and the recommendation to switch to IASI-C in October 2019 is well justified. Consider providing a concrete prescription in the data usage guidelines, e.g., “For long-term analyses, use IASI-A from July 2007 to September 2019 and IASI-C from October 2019 onward.” Users would benefit from this explicit guidance rather than having to piece it together from the discussion.

References:

Capelle, V., Chédin, A., Pondrom, M., Crevoisier, C., Armante, R., Crepeau, L., and Scott, N. A.: Infrared dust aerosol optical depth retrieved daily from IASI and comparison with AERONET over the period 2007–2016, Remote Sens. Environ., 206, 15–32, https://doi.org/10.1016/j.rse.2017.12.008, 2018.

Clarisse, L., Clerbaux, C., Franco, B., Hadji-Lazaro, J., Whitburn, S., Kopp, A. K., Hurtmans, D., and Coheur, P.-F.: A decadal data set of global atmospheric dust retrieved from IASI satellite measurements, J. Geophys. Res. Atmos., 124, 1618–1647, https://doi.org/10.1029/2018JD029701, 2019.

Di Biagio, C., Formenti, P., Balkanski, Y., et al.: Global scale variability of the mineral dust long-wave refractive index: a new dataset of in situ measurements for climate modeling and remote sensing, Atmos. Chem. Phys., 17, 1901–1929, https://doi.org/10.5194/acp-17-1901-2017, 2017.

Huang, Y., Kok, J. F., Kandler, K., Lindqvist, H., Nousiainen, T., Sakai, T., Adebiyi, A., and Jokinen, O.: Climate Models and Remote Sensing Retrievals Neglect Substantial Desert Dust Asphericity, Geophys Res Lett, 47, e2019GL086592, https://doi.org/10.1029/2019GL086592, 2020.

Saito, M., Yang, P., Ding, J., and Liu, X.: A Comprehensive Database of the Optical Properties of Irregular Aerosol Particles for Radiative Transfer Simulations, J Atmos Sci, 78, 2089-2111, 10.1175/jas-d-20-0338.1, 2021.

Zheng, J., Zhang, Z., Yu, H., Garnier, A., Song, Q., Wang, C., Di Biagio, C., Kok, J. F., Derimian, Y., and Ryder, C.: Thermal infrared dust optical depth and coarse-mode effective diameter over oceans retrieved from collocated MODIS and CALIOP observations, Atmos. Chem. Phys., 23, 8271–8304, https://doi.org/10.5194/acp-23-8271-2023, 2023.

Zheng, J., Yu, H., Zhou, Y., Shi, Y., Zhang, Z., Di Biagio, C., Formenti, P., and Smirnov, A.: A novel retrieval of global dust optical depth and effective diameter based on MODIS thermal infrared observations, Remote Sens. Environ., 332, 115083, https://doi.org/10.1016/j.rse.2025.115083, 2026.