Toward Aerosol-Aware Thermal Infrared Radiance Data Assimilation
, Emily Liu 3, Andrew Collard 3, Benjamin Johnson 2, Cheng Dang 2 and Patrick Stegmann 4,5
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis study presents a timely and relevant investigation into the development of an aerosol-aware data assimilation (DA) framework for thermal infrared observations, specifically using IASI radiance data. Given the increasing importance of all-sky and all-condition assimilation in modern numerical weather prediction (NWP), the effort to account for aerosol effects—particularly their impact on upwelling radiance in the IR window—represents a meaningful step forward.
The manuscript is clearly written, logically structured, and addresses a technically challenging problem. The proposed DA framework introduces two key components: (1) detection of aerosol-affected pixels, and (2) an observation error model that incorporates the magnitude of aerosol effects. These are well-motivated and demonstrate a thoughtful approach to integrating aerosol information into the assimilation system. However there are some suggestions:
- More details on how the aerosol detection algorithm was implemented should provided in the paper. For instance, what specific spectral channels or features were used? How sensitive is the detection to different aerosol types (e.g., dust vs. smoke)?
- Clarify how the aerosol optical depth (AOD) or other proxies were obtained for constructing the error model. Were they from external sources (e.g., CAMS), or derived from the observations themselves?
- While the paper reports reduced SST biases, it would be helpful to include quantitative metrics such as root-mean-square error (RMSE) or correlation coefficients before and after assimilation to better quantify the improvements.
- The statement that the analyses and forecasts show “neutral results” should be supported with more detailed evaluation statistics across multiple cases or time periods.
- The conclusion mentions future improvements related to quality control and bias correction. It would strengthen the manuscript if the authors could elaborate on possible strategies for addressing these issues—for example, whether machine learning approaches might be explored for improved aerosol detection or dynamic error estimation.
- Was there any computational overhead introduced by the aerosol-aware framework? A brief discussion on the efficiency and scalability of the method within the operational DA system would be valuable.
- Some sentences could be rephrased for clarity and grammatical correctness (language editing is recommended).
- Ensure consistency in terminology (e.g., "hazy-sky" vs. "aerosol-affected" conditions).
Overall, this is a solid and promising contribution to the field of data assimilation, particularly in the context of handling non-clear-sky and non-standard atmospheric conditions. With minor revisions and additional technical details, the manuscript will be suitable for publication.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsPlease see the attached file.
Comments for author File: Comments.pdf
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsI recommend the Authors focus on the following aspects to improve the scientific quality of their manuscript.
1) Table 1 shows that with the reduced NCEP channel set, ADM misidentifies a large fraction of pixels (many “hazy-sky” pixels have zero LMDAERO AOD and vice versa). Consider using the full IASI spectrum (or additional spectral indices) to better capture dust and smoke signatures, or adapt thresholds regionally.
2) ADM is applied only over water here, yet many operational applications would require robust detection over land (and for different aerosol types). A more extensive, quantitative comparison against independent AOD retrievals (e.g., VIIRS level-2/3, ground-based AERONET) would demonstrate where and when the detection works (or fails).
3) It appears that cloud QC rejects many legitimately dust-laden scenes (Figure 6), which may bias the analysis toward only moderate loading. A sensitivity study showing how relaxing vs. tightening these QC filters affects sample size and analysis skill would clarify the trade-off between purity and coverage.
4) The framework ingests MERRA-2 mass mixing ratios as “truth” yet real aerosol plumes often exhibit vertical layering that reanalyses may not capture. A set of single-column tests (or full 4DVar sensitivity experiments) in which the peak altitude and thickness of the dust layer are perturbed would reveal how uncertainties in vertical placement propagate into BT simulations, Jacobians, and ultimately the meteorological analysis.
5) As shown in Wei et al. (2022) and in Section 3.2, aerosols alter Jacobian sensitivities differently for T and q. A more explicit breakdown of how Ae–dependent error inflation (Section 3.3) affects temperature and moisture increments would help readers assess the framework’s physical consistency.
6) The current approach excludes aerosol-flagged data from bias estimation, yet then applies those clear-sky-derived biases to the hazy-sky radiances. A simple Ae-dependent linear correction (Figure S6) overcorrects SST, suggesting that existing predictors (e.g., scan angle, surface type) are insufficient. I recommend exploring additional predictors—such as Ae itself, modeled AOD, or BTDs—that might capture the systematic radiance offsets induced by aerosols without degrading clear-sky performance.
7) The two-month “observer” experiment focuses on Saharan dust; however, smoke and mixed aerosol events behave differently spectrally. Including at least one contrasting period (e.g., biomass-burning season) in the bias-correction derivation would show whether the proposed model generalizes beyond mineral dust.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
Comments and Suggestions for Authors**General comments.**
It would be beneficial for readers from the data assimilation community to have access to a concise section with a general overview of the data assimilation technique used, as well as a more detailed commentary with additional specific information of this part crucial to the methodology. For example. the present study investigates the implementation of a 4DEnVar technique, and would be interesting to see is whether the solution was handled by way of a parameters or state perturbation in the analysis update problem. A further research question is whether the emissions or the concentrations were updated.
The title of this manuscript, "Towards aerosol-aware thermal infrared radiance data assimilation", suggests the implementation of a process that has not yet been achieved. Might this be the first paper in a series of papers that will achieve this objective?
Minor **comments**:
Line 20 "shows reduced biases" --> "reduces biases"
Line 22 Clarified what "neutral results" means
other suggested keywords: aerosol; radiance data assimilation; thermal infrared; IASI; all-sky assimilation; satellite retrievals; observation error modeling; aerosol detection; sea surface temperature bias; dust plumes; infrared sounders; atmospheric composition; numerical weather prediction (NWP)
Line 28 "through its transmittance effects" --> "through their transmittance effects"
Line 30 " as a form in brightness 30 temperature (BT), [1-5]." -> as a form in brightness 30 temperature (BT)[[1-5].
Line 37 "top of the atmosphere in model." -> top of the atmosphere in the model.
Line 45 "directly assimilate the IR measurements with ingesting the modeling aerosol information" → "directly assimilated IR measurements by incorporating modeled aerosol information"
Line 48- "they also reported substantial changes to the quality control and bias correction" → "they also reported substantial changes in quality control and bias correction"
Line 62- "to identify the IR measurements affected by aerosols (i.e., hazy-sky)" → "to identify IR measurements affected by aerosols (i.e., hazy-sky conditions)"
Linea 75- "Underutilizing of considerable volume" → "The underutilization of a considerable volume"
Line 91 - "implemented at 2021" → "implemented in 2021"
Line 106 - "range from ultraviolet to microwave" → "ranging from the ultraviolet to microwave spectrum"
Line 112 - "manipulate the simulations" → "perform the simulations" or "conduct radiance simulations"
Line 131 - **"Dataset" (as section header) → "Datasets"**
Plural, since multiple datasets are described.
Line 153 - **"used subjectively to compare with MERRA-2 simulation" → "used qualitatively to compare with MERRA-2 simulations"**
"Subjectively" is vague in a scientific context; "qualitatively" is more appropriate.
Line 161 - "Climate Date Store" → "Climate Data Store"
Line 165 - **"an aerosol-aware framework proposed in this study consists of the following items"** → _"The aerosol-aware framework
proposed in this study consists of the following components"_
Line 174 - **"as part of first-guess fields"** → _"as part of the first-guess fields"_
Line 182- **"The observation error will increase when A_e is larger, vice versa."** → _"Observation error increases with A_e and decreases as A_e diminishes."_ (“Vice versa” is vague and informal here.)
Line 187 - **"More detail information can be found"** → _"More detailed information can be found"_
Line 193 - Just to confirm, Figure 2 says that Color level shows in logarithm scale. Is it right in a log scale?
Line 204 - **"It could lead to mis-identifications"** → _"This may lead to misidentifications"_
Line 209 - **"more light loading pixels"** → _"more pixels with light aerosol loading"_
Line 214 - **"73% v. 86%"** → _"73% vs. 86%"_
(“vs.” is standard and more widely understood.)
Line 216 - **"cloud-contamination"** → _"cloud contamination"
Line 232 - **"Combination of these two terms, A_e symmetrically quantifies"** → _"The combination of these two terms symmetrically quantifies A_e"_
Line 235 - **"A_e shows strong negative signal contributed by both terms"** → _"Ae shows a strong negative signal contributed by both terms"_
Line 250 - **"rejects the observation where the absolute first-guess departure"** → _"rejects observations where the absolute first-guess departure"_
Line 261 - **cannot capture the condition involving the misidentification in ADM"** → _"cannot capture cases where ADM misidentifies cloud-contaminated observations as hazy-sky"_
Line 347- **"in AER experiment"** → _"in the AER experiment"_
Line 347 - **"as a function of assimilated IASI channels"** → _"as a function of the assimilated IASI channels"_
Line 351 - **"smaller magnitude of cooling effect compared to [20]"** → _"smaller magnitude of the cooling effect compared to that reported in [20]"_
Line 352 - **"due to no stratification of aerosol optical depth"** → _"due to the lack of stratification by aerosol optical depth"_
Line 408 - **"it shows a similar story"** → _"the results follow a similar pattern"_
Line 359 **"in AER"** → _"in the AER experiment"_
Line 360 - **"over trans-Atlantic region and Arabia Sea"** → _"over the trans-Atlantic region and the Arabian Sea"_
Line 422- Figure 8. --> Distribute better within the free space of this mosaic.
Line 497 and Line 510 - considered put to this not shown Consider: _"(figure not shown)"_
Line 577 - -**"comes to the case"** → _"is fully implemented"_
**References**
Include some recent references (2025) and more from 2024 to refresh the state of the art of the current manuscript.
The manuscript is a well written paper that provides a methodology for data assimilation of satellite radiances to update through the analyses the meteorological driver. The methods content with the specific details of the experiment is supported for easy lecture and consult in the supplementary material.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThe revisions have been completed as requested, and the manuscript is approved for publication.
Reviewer 3 Report
Comments and Suggestions for AuthorsI believe that the Authors did a good job in revising their manuscript that is now ready for publication in my opinion.
