the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 25 Aug 2026
Unified Aerosol Algorithm: New insights into the global aerosol system from the Ocean Color Instrument on NASA's PACE mission
Abstract. The launch of NASA's Plankton, Aerosols, Clouds, ocean Ecosystem (PACE) mission in 2024 with the Ocean Color Instrument (OCI) on board created new opportunity for characterizing aerosol properties from space. OCI observes Earth and its atmosphere across a broad spectral range in the reflective spectrum from the ultraviolet (UV) to the Short Wave InfraRed (SWIR). Heritage aerosol retrieval algorithms developed to accommodate radiometers measuring in the visible to SWIR wavelengths such as the MODerate resolution Imaging Spectroradiometer (MODIS) have been successful in characterizing aerosol optical depth (AOD) and indications of aerosol particle size but are less sensitive to aerosol particle absorption or aerosol layer height. Other heritage sensors measuring in the UV part of the spectrum such as the Total Ozone Mapping Spectrometer (TOMS) or the Ozone Monitoring Instrument (OMI) are sensitive to absorption and layer height but have insufficient information to constrain AOD in an aerosol retrieval. With OCI encompassing the entire spectral range of interest, the Dark Target and Deep Blue algorithms from the MODIS tradition and Near UV algorithm from the TOMS/OMI tradition are brought together, adapted for OCI and unified for retrievals of AOD, particle size parameter, aerosol absorption and aerosol layer height. Adaptations for OCI include adjustment for OCI sensor characteristics, modification of cloud and snow masking routines, production of the traditional UV Aerosol Index at unprecedented 1 km resolution, merging of the Dark Target and Deep Blue retrievals over land, extrapolation of spectral AOD retrieved in the visible range into the UV range, use of Oxygen-B bands for aerosol layer height, and retrieval of AOD above clouds that provides visualization of All Sky aerosol loading, Six months of retrievals have been compared with collocated AERONET AOD and single scattering albedo (SSA) resulting in a preliminary validation. AOD is biased high for shorter wavelengths, especially over ocean leading to asymmetrical error bounds for AOD. No similar bias is seen in the SSA retrievals which are exhibiting error bounds of ±0.04, at this stage. The results of the OCI Unified Aerosol Algorithm are quantitative characterization of the global aerosol system over ocean, vegetated and barren surfaces, in clear skies and above clouds.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Atmospheric Measurement Techniques.
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
(3863 KB) - Metadata XML
-
Supplement
(869 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-3731', Anonymous Referee #3, 16 Sep 2026
-
RC2: 'Comment on egusphere-2026-3731', Anonymous Referee #2, 19 Sep 2026
This manuscript provides the description of retrieval algorithm of PACE OCI (including its heritage connecting to previous well-known algorithms), and some preliminary evaluation results of new products. It is generally well-written and very informative, therefore it looks fine to be published. Although this manuscript itself can suggest the useful ideas to the research community, it will be much better after some revisions. Please do a revision based on the suggested comments as below.
- UAA (unified aerosol algorithm) is one of keywords in this manuscript, I think. But this word was not included in the abstract. I think that it is necessary to include some statements about the UAA. Also, a writing 'UAA algorithm' is wrong. Please correct it 'UAA'.
- In Fig. 2, the meaning of 'bias_diff' should be written in the figure caption.
- Fig. 3 shows the case study of Alta Floresta. Is there any special reason to show this site, or results in other sites are similar each other, so the case in Alta Floresta is simply shown as an example? Since this manuscript is not the special case study, it may be better to use mean pattern when a certain result is provided. If not (i.e., a certain site result is provided), please clarify that it is an example, and other cases show similar pattern
- What is the role of Fig. 5 in this manuscript?
- In Figs. 10 and 11, I do not figure out the necessity to investigate the aerosol layer height (ALH). What can we learn from this analysis? I think that the meaning of ALH is not well discussed in the research field. It will be very helpful if authors can add meaningful interpretation of this ALH pattern combined with the analysis of other results. It can be a good example to show how to use this ALH information in the study of satellite aerosol products.
- Fig. 12 shows the case in August 2024. Why this time period is selected? Is there any special reason to see the case of August 2024 as the example?
- In Figure 13, probably a panel figure was not included related to the result of 870 nm. Please check again.
- Still, the quality of SSA is OK just in the case of AOD > 0.4? The usage of SSA was suggested under this condition (AOD > 0.4) about in 1990s, and this limitation is still maintained after about 30 years. Is there any possibility to get over this limitation? The condition 'AOD > 0.4' is actually strong, therefore the application of SSA to the general research is still very difficult. Is there any possibility to solve this problem based on AI tools or comparable ways? I would just ask the idea / insight of authors who are very professional about this issue.
Citation: https://doi.org/10.5194/egusphere-2026-3731-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 223 | 99 | 48 | 370 | 61 | 51 | 48 |
- HTML: 223
- PDF: 99
- XML: 48
- Total: 370
- Supplement: 61
- BibTeX: 51
- EndNote: 48
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
1. Section 1: Most statements in the Introduction are supported by only a single reference. Because this article is combination of three representative heritage algorithms, the references would be expanded so readers can trace the lineage of each component.
2. L90–92. The phrase of "two pieces of information are available in the UV" is reasonable, but needs supporting references.
3. L96. "Equally weighted" is not the best description. The sensitivities of aerosol amount, optical absorptivity, and layer height are not equal; they are mutually dependent. Please reword, e.g. "the three parameters strongly influence one another."
4. L127–132. This part does not fit the NASA heritage framing of the surrounding text, particularly the GOSAT-2/CAI-2 material. I suggest removing it.
5. L169. The meaning of "0.005 µm bands of 0.0025 µm spectral steps" is unclear. Do the authors mean a spectral resolution of 0.005 µm with a sampling interval of 0.0025 µm? Please state this in standard terminology.
6. L191. Algorithm references should be given here.
7. L211–213. Is η, defined as the fine mode fraction at 0.55 µm, applied at all wavelengths? If so, the spectral dependence of the fine mode fraction is not represented. Please comment on the resulting error, especially for the UV extrapolation over ocean.
8. Sections 4 and 5. The split between flow (Section 4) and innovations (Section 5) forces readers to move back and forth, since Section 4 repeatedly forward-references Section 5. Merging the two would improve readability.
9. Section 5.1.
1) Please confirm that the VIIRS LUTs are used with no modification. Differences in band centers and widths between VIIRS and OCI should introduce some inconsistency; was this quantified?
2) The loss of thermal IR is treated qualitatively only. Residual cloud contamination is a plausible cause of the positive AOD bias at low AOD. Cloud fraction is already in the output files, so stratifying the AOD bias by cloud fraction would settle this.
10. Section 5.3. The merge uses fixed latitude–longitude boundaries, which will produce discontinuities at the boundaries. Please quantify the AOD step across representative boundaries and discuss the impact on regional means and future trend analysis.
11. Section 5.4, Figs. 3 and 4. The 80/20 split is random, so collocations from the same site and day appear in both sets. With site location and month as model inputs, the reported MAE and RMSE may reflect memorized site climatology rather than extrapolation skill. Can the authors repeat the evaluation with a site-wise hold-out, and does the accuracy hold? The same concern applies to Fig. 3 if Alta Floresta was in the training set.
12. Section 5.5. The advantages of the near-UV algorithm are described, but no sensitivity analysis for SSA or ALH is given. Section 5.7 provides exactly this kind of analysis (Figs. 7, 8) for the O2-B method, so the contrast is noticeable. Please show how the near-UV signal responds to SSA and ALH with AOD fixed, and how an AOD error propagates into SSA and ALH.
13. Section 6. Validation coverage is uneven. AOD and NUV SSA are evaluated against AERONET; the two ALH products are not evaluated at all.
It is also not explained why two ALH products from different spectral regions are provided, or what each actually measures. Figs. 10 and 11 show large disagreement, and the dust case gives essentially different fields, but no inter-comparison statistics or discussion of causes are given.
This matters because L961–963 recommends using AOD, SSA, and ALH together for assimilation and flux calculations. Please add a quantitative inter-comparison, discuss the likely causes, and give explicit usage guidance. The authors already apply a clear standard to the DT SSA by declaring it a diagnostic; the same standard seems to apply to ALH.
14. Fig. 13. Near-UV retrievals appear largely absent below AOD ~0.1, unlike the visible. Is this an explicit threshold (the AOD > 0.2 condition at L941?) or a convergence failure? If it is a threshold, it should be stated in Section 5.5, with a comment on the sampling bias introduced by applying a threshold to a positively biased AOD.
Technical corrections