the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 09 Jul 2026
Implementing Marine Aerosol Ice Nucleation Parametrizations in the Unified Model: Mitigating the Cloud Radiative Bias During a Case Study Over the Southern Ocean
Abstract. Mixed-phase clouds over the Southern Ocean profoundly influence Earth's radiative balance, yet climate models persistently exhibit a positive surface shortwave radiation bias driven by over-glaciation in mixed-phase clouds. Addressing this bias requires accurate representation of ice-nucleating particles (INPs). This study presents the first online implementation of aerosol-aware marine INP parametrizations derived interactively from sea spray and marine organic aerosols within the high-resolution Unified Model. Using CAPRICORN-2 shipborne observations, we demonstrate that the default INP scheme overestimates INP concentrations by up to four orders of magnitude, causing a severely underestimated liquid water path. In contrast, empirical Antarctic and deterministic marine INP schemes reproduce the low INP concentrations typical of the pristine Southern Ocean, improving cloud and radiative properties. Crucially, microphysical and radiative responses to these INP reductions are strongly regime-dependent. In deep, pre-frontal mixed-phase clouds, suppressing INP concentrations effectively inhibits cloud glaciation, significantly enhancing supercooled liquid water and largely mitigating the surface shortwave radiation bias. However, in shallow, post-frontal stratocumulus clouds, altering INP parametrization yields negligible improvements. Radiosonde evaluations reveal this insensitivity is driven by several model deficits, including systematically smoothed boundary layer inversions leading to excessive dry air entrainment, exacerbated by underestimated cyclonic moisture transport. Consequently, these shallow clouds are thermodynamically starved of water vapor, rendering microphysical INP adjustments ineffective. Ultimately, fully resolving Southern Ocean cloud-radiation biases requires synergistic advancements in representing aerosols, boundary layer physics and large-scale meteorological forcings.
- Preprint
(16356 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 14 Nov 2026)
- RC1: 'Comment on egusphere-2026-3541', Anonymous Referee #1, 21 Aug 2026 reply
-
RC2: 'Comment on egusphere-2026-3541', Anonymous Referee #2, 23 Oct 2026
reply
General appreciation
This research manuscript tackles the question of radiative bias of modelled clouds over the Southern Ocean, by implementing and evaluating parameterizations of ice nucleation in a high-resolution model. The study focuses on a case study motivated by the observational campaign CAPRICORN-2 that took place in early 2018. The authors used the coupling of an aerosol scheme (GLOMAP-mode) and a double moment microphysics scheme (CASIM), to implement online aerosol-to-INP parameterizations, that are then compared to more classic temperature-based ice nucleation parameterizations. The study provides insights on the sensitivity of modelled cloud microphysics to INP representation, that appears to be a second order determinant for the model radiative bias. Namely, ice nucleation parameterizations yield to significant divergence in cloud properties only during mixed-phase cloud events, and do not explain the main model-observation biases.
This works provides a deep analysis of the ice nucleation parameterizations’ effects on the representation of cloud phase over the Southern Ocean. Its results shed light on both the role of marine aerosols for ice nucleation, and the sensitivity of modelled cloud phase to ice nucleating particles number concentration. The comparison methods are acknowledgeable, thanks to the use of remote sensing simulator for model output, and to the clear testing protocol. However, comparisons often lack quantification that would strengthen the results’ scientific reach. In addition, the key scientific outputs are blended in extensive descriptions of figures that tend to be redundant, to the detriment of a deepened investigation of the physical mechanisms identified, in the paper, as the cause for the observed model biases. Nevertheless, the gradual narrowing of the analysis, from a broad comparison of cloud phase representations among the model setups, to a closer and more detailed look at the effect on radiative fluxes, helps the reader to understand the implications of ice nucleation modelling on cloud radiative budget.
For the reasons mentioned hereinabove, I would recommend this paper for publication after corrections that follow the comments below.
Major comments:
- The description of the results is too long and too descriptive. I recommend shortening it by relying, as much as possible, on relative rather than absolute comparisons (e.g. fig. 4, table 1). This would make the quantification of the highlighted differences between models and observations easier, and thereby would make the conclusions clearer.
- Figure 9 does not provide quantitative information on the model comparison. It is an nth comparison of the parameterization setup outputs, that only highlights the fundamental difference between parameterization C86 and the others; difference already highlighted several times in the previous figures, starting from Fig. 3. The only added value of the figure is that it shows that this discrepancy has consequences only in mixed-phase scenarios. Therefore, I suggest revising this figure, by either relegating it to the appendix section, or making it more specific about the difference it shows thanks to an extra panel that would display the quantification of this difference.
- The discussion about aerosol and parameterization uncertainties have too much importance in the discussion section, considering the conclusions about the insensitivity of cloud and radiation to INP concentration changes, the latter being identified as second order drivers. The discussion should rely on this important result to focus on the possible physical processes that drive the model biases.
The manuscript already contains these elements of discussion, but they need to be highlighted and brought above the details, which can be shortened, about the second order drivers.
Minor comments:
- L36-37: the sentence seems not grammatically correct, or unclear.
- L34 to 41: you can cite the work of Raillard et al. 2025 on the development of turbulence informed cloud phase parameterization, evaluated over the SO (DOI: 10.1029/2025MS005100)
- L49: please mention the other nucleation mechanisms.
- L52: please provide a reference for the affirmation: “marine aerosols become the primary driver of ice formation”.
- L95: “vital” could be replaced by a more proper term for relating to a physical process.
- L164: you mention Fig. 1, where one can notice that part of the ship track is close to the domain edges. Can you please demonstrate that the comparison between observational data and model outputs does not suffer of domain boundary effects for the points close to the domain edge?
- L174 to 176: please provide some quantification for the preliminary testing you mention. Even though these evaluations are outside the scope of the present paper, it would be appreciated to have a supplementary figure that illustrate the effects of the corresponding changes.
- L190-191: please be more specific about how GLOMAP-mode explicitly resolves the aerosol size distribution. It is not clear if it uses a size-bin representation, or a continuous spectrum of size. Either way, it needs to be mentioned and shortly described in the paragraph.
- L220: The sentence “Originally derived from airborne measurements.” should be removed or incorporated in the previous sentence.
- In section 2.2: I suggest tagging the INP parameterizations as aerosol-aware and non-aerosol-aware. It would highlight this fundamental difference to the non-specialist reader. In the following (L360), you refer to W15 and M18 as dynamical schemes. I suggest replacing this term by aerosol-aware, and/or to introduce it before in the text.
- L339: please specify what conversion you talk about. If it is from number per cubic meter to number per liter, the sentence can be removed for readability.
- L341: you indicate a temperature range from the warmer to the colder, while in the following you use the reverse convention, i.e. from the colder to the warmer. Consider adapting for consistency.
- L381-382: here, the reader’s attention is drawn to Fig. 4, and they are encouraged to compare the model outputs (panels b-e) with the observations (panel a). However, it is fastidious to perform that comparison with the naked eye on Fig. 4. The figure would benefit a sixth panel that pinpoints the time-height locations where the model differs from the observations. Since the four parameterizations show similar structures, I suggest that the difference is computed with an “average” of the models. The average could be produced by tagging the phase with numbers (clear=0, liquid warm=1, ice=2, and so on) and then rounded to the nearest integer.
- Figure 7: the line associated to V21 is barely visible. Please change its color or find a way of making it more visible.
- L430-432: the statement described here is very important: all model configurations have the same bias in moisture vertical distribution. Please consider adding a sentence that highlights the little role of ice nucleation parameterizations, and that suggests another mechanism responsible for this bias.
- L609-612: This should be in a “Limitations” section, along with some of the elements about the aerosol source and parameterization uncertainties.
- L666-672: The recommendations proposed here should be clearly identified as first order and second order drivers, rather than presented as a list with no clear importance ranking.
- L680-681: In the statement, data visualization code is mentioned but the repository does explain how to use it. Please add in your Zenodo repository, a more extensive description of the shared data and code, and a README about how to use it.
Citation: https://doi.org/10.5194/egusphere-2026-3541-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 413 | 184 | 66 | 663 | 44 | 60 |
- HTML: 413
- PDF: 184
- XML: 66
- Total: 663
- BibTeX: 44
- EndNote: 60
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This study focuses on applying the UM regional model to simulate a Southern Ocean frontal clouds case to analyze the pre and post front mixed phase clouds with different parameterizations of Ice Nucleating Particles (INPs), with evaluations carried out using the CAPRICORN observation data. I suggest major revision for this current draft and some parts of the manuscript are too general for me to provide specific comments, but I am happy to review its second round.
This study shows comprehensive variables including hydrometer type, liquid/ice water path, temperature and humidity vertical profile, surface downwelling shortwave/longwave radiation, INPs number, surface meteorologies, etc. Technically, the analysis is cloud-phase microphysics oriented, and the focus on time series study intended to illustrate the development of the supercooled liquid clouds/mixed phase clouds as a system with simulated INPs as the changing factor. The strongest selling point of this paper, in my personal point of view, is scientific cross-validation, including the knowledge of potential related bias sources diagnoses and the comprehensive illustration of model performance using 2 regimes. The weak point of this paper, however, is its novelty and the depth of study.
To be a little specific, I acknowledge the credit of the research topic, methods, choice of model input, simulator, and most of the statistics shown in this paper. However, the novelty and depth of this study is expected to be further explored in the next round of review. For example, “the first online implementation of sea spray and marine organic aerosols as INP sources…”, the study didn’t fully explain the advantages of 1) online implementation of INP parameterization and 2) dynamical aerosol-aware parameterizations under this particular case. There indeed shows the various INP numbers and simulated hydrometers using different INP parameterizations as a result, and their accompanying statistics shows maybe a better general performance. But the authors didn’t show/ didn’t clearly illustrate the reason behind it. Similarly, I have a feeling, from both the introduction section and the description of figures, that immersion freezing should be the focus of your microphysics study. Again, the to-the-point diagnosis hasn’t been fully illustrated to tell the story out, which in theory would exactly be this high resolution UM RAL model’s advantage.
Similarly, the physics between the thermodynamics profiles, microphysics influenced by the INP schemes, and the resulting radiation fluxes are not dynamically connected. The hints have been listed all through the manuscript and the variables are reasonably chosen. Had this study been a simulation-observation class project final report, I would rate highly for its hard working, fruitful results, and comprehensive statistics. But for journal publication, I personally would like to see by changing the cloud liquid water into rain/ice crystals/snow/graupels, how do the INPs specifically indirectly influence the long/shortwave radiations. For the MIX period and SLW period specifically, what type of biases of Temperature and Humidity structure, adding in what extent of INP bias, by location or/and by population, resulted in how much cloud liquid water transfer into ice crystals/rain/graupel/supercooled liquid water through what processes? What are the above contributions from sea spray INPs and what are the corresponding results from the marine organic INPs. Did the sea ice representation affect the available water vapor source? Are the front location and low pressure center location and intensity well captured/nudged by the model/ERA5? Except for the hydrometer type, are the cloud cover/fraction represented well by the model in general with different INP schemes? I am happy to provide more specific comments in the future rounds of reviews when more in depth analysis is illustrated.
Minor:
Abstract can be more quantitative: try to avoid using confusing terms such as “default”, “high-resolution”,”starved of water vapor”. What is the default, explicitly mention the spatial resolution, quantify the underestimation of water vapor, suppressing INP number effectively inhibit cloud glaciation through what mechanism, etc..
Some statements have citations which are related and beneficial, but didn’t include the very original paper.
Some paper reviews can be more Southern Ocean oriented.
Suggest rephrasing Section 2.4, adding some concise and clear information about the lidar-radar merged products and the Silber et al. 2022 simulator and related hydrometer cluster method, just for reader’s convenience. Details can be listed in the Appendix. Similar for Section 2.2.4, maybe update this section with a Table for better vision and leave details in the Appendix. Similar for the model configurations, if a Table can better tell the input and schemes. The ultimate goal is to show the physical story line more clearly.
Suggest shortening the discussion part about potential uncertainties, instead focus on what the current data express with significance.