the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 25 Feb 2026
The MISR Research Product Algorithm – Producing Global, Self-Consistent, Pixel-Level Aerosol Retrievals for the Multi-angle Imaging SpectroRadiometer
Abstract. The MISR Research Aerosol retrieval algorithm (RA) was developed in parallel with the MISR standard operational algorithm (SA), to explore innovative retrieval ideas for possible implementation in the SA, and to provide higher-spatial-resolution results with enhanced sensitivity to particle properties, where possible. Whereas the SA was designed to run automatically on the entire MISR dataset, the RA, a research code, had to be run on a case-by-case basis, with considerable user involvement. We present here a version of the RA, the MISR Research Product Algorithm (MRPA), that has been streamlined and automated for wider use, along with validation results for this algorithm. The compromises required to automate the RA result in somewhat diminished particle-type sensitivity. However, the MRPA provides 1.1 km pixel resolution (compared to 4.4 km for the SA), and comparing the AOD data statistically with AERONET, it offers about half the statistical RMSE, significantly higher correlation coefficient, though somewhat higher bias, for mid-visible AOD compared to the SA over land, along with better Ångström exponent statistics, especially at higher AOD. Statistical validation of particle shape and single-scattering albedo is more difficult, but for sufficiently high AOD, retrievals in smoke-dominated regions show a preponderance of small-medium, spherical, light-absorbing particles, whereas dust-dominated regions tend to have larger, weakly absorbing, non-spherical particles.
- Preprint
(5450 KB) - Metadata XML
-
Supplement
(3574 KB) - BibTeX
- EndNote
Status: open (extended)
-
RC1: 'Comment on egusphere-2026-342', Anonymous Referee #1, 14 Apr 2026
reply
-
AC1: 'Reply on RC1', Michael Anstett, 06 Jun 2026
reply
We thank the reviewer for constructive and very helpful notes, and we have done our best to address each point, as summarized below. Proposed additions to the manuscript and supplement in response to comments such as figures and tables have been isolated in the attached supplement.
R.1 It is not easy to assess specifically whether the revised approach reduces sensitivity to calibration residuals. However, as MISR radiometric calibration issues are generally largest for the blue band, this might be mitigated to some extent.
Regarding the last point, it is possible that the revised mixture weighting is overly aggressive. But overall, the MISR aerosol-type retrieval amounts to selecting from among the mixtures in the algorithm climatology those that most closely match the observed radiances, so by far the most dominant factor affecting how well the retrieval results represent the actual atmospheric state is the assumed climatology. For this reason, we have aimed to present the limitations as well as the strengths of our particle-property retrieval results.
R.2 An analysis of algorithm performance for three regimes of scene-wide NDVI has been added, comprising Table S4 and remarks in Sect. 3.1 (Contained in the supplement to these comments). Ideally, such analysis would be done based on pixel-level NDVI, though unfortunately this is not feasible for us currently. As it stands, we hope that Table S4 illustrates the rationale acceptably.
R.3 The added Table S5 contains statistics for the three component algorithms, constrained for PSA AOD > 1.0. Here, the PSA is used as it determines the mixing coefficients for the CSA. The value of the PSA mixing may not immediately obvious from the RMSE and MAE, but the lower absolute bias compared to either the RSA or PSA should help demonstrate the usefulness of this method.
R.4 This is a persistent challenge for satellite aerosol retrievals in general. We do not know of any dataset that provides systematic ground-truth for such an assessment. The reviewer has summarized the (qualitative) rationale for the added effort we introduced, with the aim of improving the cloud masking, but absent a ground-truth dataset, we are not sure how to take the assessment further.
R.5 A supplemental analysis has been added in Sect. 3.4, comprising of Angstrom exponent, Single-scattering albedo, and fine-mode fraction plots and statistics for MISR AOD > 0.2, performed on a selection of sites in meteorological winter. This selection of sites is detailed in Table S2.3.
R.6 We don’t have a way of getting the Garay et al. (2020) cases specifically, and regenerating such a dataset would be a considerable amount of work for what seems likely to be very limited benefit.
-
AC1: 'Reply on RC1', Michael Anstett, 06 Jun 2026
reply
Model code and software
MISR Research Product Algorithm Michael Anstett and James Limbacher https://github.com/mranst/misr_research_product_algorithm
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 689 | 291 | 65 | 1,045 | 137 | 72 | 127 |
- HTML: 689
- PDF: 291
- XML: 65
- Total: 1,045
- Supplement: 137
- BibTeX: 72
- EndNote: 127
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review of “The MISR Research Product Algorithm – Producing Global, Self-Consistent, Pixel-Level Aerosol Retrievals for the Multi-angle Imaging SpectroRadiometer”. This article presents substantial updates to the MISR research algorithm, and analysis of how the updates compare with AERONET and other MISR products. Overall, the paper demonstrates the value of the product. I look forward to seeing this paper successfully published, and believe it will contribute meaningfully to the academic discourse. To help the paper to achieve its best results, I have six questions that I would like the authors to address. I believe that addressing these minor/moderate changes will make the paper even more successful.
Q.1. NIR-only Bc and revised mixture weighting (methodological refinements). The shift to NIR-only computation of the shape-similarity coefficient Bc (Eq. 2) is physically well-motivated, as NIR radiation penetrates the atmosphere more effectively and is less susceptible to the flat-field anomalies and stray-light effects that are most pronounced in the blue and green bands. Similarly, the revised mixture-weighting scheme (Eq. 3b) sharpens aerosol type selection by concentrating weight on the lowest-cost mixtures. Could the authors briefly comment on whether these two changes also reduce sensitivity to calibration residuals or cloud-contamination artifacts? Additionally, could the revised mixture weighting be overly aggressive in genuinely mixed aerosol scenes - for example, dust-smoke mixtures over West Africa - where a broader weighting might better represent the actual atmospheric state?
Q.2. NDVI-based CSA weighting and performance over bright/semi-arid surfaces. The updated over-land CSA weighting (Eq. 6), which incorporates NDVI thresholds in addition to PSA AOD, is well-justified and should help mitigate retrieval errors over bright desert surfaces. Do the authors have any quantitative assessment of how much this new weighting improves AOD retrievals specifically over semi-arid or bright-land surfaces relative to the Limbacher et al. (2022) version? In particular, the transition zone (NDVI 0.10-0.25) is where the blending is most active - a brief stratification of validation statistics by NDVI bin (low/medium/high) in the supplement would substantially strengthen this result.
Q.3. The authors highlight the PSA's role in correcting the persistent low bias at high AOD that affects both the SA and earlier RA versions (lines 55-62). This is an important methodological contribution. Could the authors provide, either in the main text or supplement, a more explicit quantification of bias reduction specifically at high AOD (e.g., AOD > 1.0) attributable to the PSA component of the CSA? This would make the contribution of the prescribed-surface approach more transparent and easier to compare with future algorithms addressing the same problem.
Q.4. The new cloud masking scheme is notably more sophisticated than in previous RA versions, combining MAIAC land masks, MODIS-Terra brightness temperature, and MERRA-2 sea-surface temperature. However, no direct validation of the cloud mask itself is presented - there is no assessment of cloud contamination rates or false-positive masking rates, particularly over bright desert surfaces where the brightness temperature test may be less reliable. Could the authors provide at least a qualitative discussion of the cloud mask's performance, or point to any existing validation of the MAIAC cloud mask over the relevant surface types?
Q.5. The authors rightly flag the inherent limitations of AERONET almucantar inversion variables - particularly FMF, SSA, and sphericity for quantitative validation (lines 182-190). The results show for AOD exceeding roughly 0.3-0.4, non-dust scenes are almost entirely dominated by small, spherical particles, while dust scenes show the expected coarse, non-spherical signature. This finding has clear relevance for anthropogenic aerosol studies, since urban and industrial pollution aerosol shares the same microphysical fingerprint as smoke in these retrievals. Would the authors consider adding a brief supplementary analysis at a small number of urban or industrially-polluted AERONET sites (e.g., Kanpur, Beijing, or a European site) during high-pollution episodes? This would explicitly demonstrate the MRPA's utility for anthropogenic aerosol characterization without requiring major additional development, and would complement the already-strong smoke and dust analysis. Additionally, it is reassuring to note that the retrieved mid-visible SSA under high AOD conditions (roughly 0.85–0.95) is consistent with the expected range for moderately absorbing aerosol types derived from surface observations, which supports the physical plausibility of the MRPA's particle absorption retrieval in optically dense plumes.
Q.6. The authors acknowledge that the CSA and V23 SA comparison datasets are not identical, which limits the strength of the performance comparison. Would the authors consider adding a supplementary table reporting CSA statistics computed over the exact same coincidences used in the Garay et al. (2020) SA validation, to facilitate future meta-analyses and more rigorous algorithm intercomparison? Additionally, two minor points warrant checking: (a) Figure 1 and Figure 5 captions both reference "October 2023" whereas the stated study period ends October 2022 - please verify; and (b) the over-water correlation coefficient r = 0.093 reported for the CSA (Table S6) appears inconsistent with the strong visual agreement in Figure 5 and may be a typographical error.