| 19 Nov 2025
Secondary ice production affects tropical convective clouds under different aerosol conditions
Abstract. Secondary ice production (SIP) modulates tropical convection, but its interactions with aerosol loading remain uncertain. Using the UK Met Office Unified Model with detailed SIP parameterizations, this study investigates how multiple SIP mechanisms affect a Hector storm under weak instability and varying cloud condensation nuclei (CCN) concentrations. At moderate aerosol loading (Nd = 400 cm–3), SIP substantially enhances the ice phase. Ice number increases by over 3 orders of magnitude at temperatures warmer than –10 ℃, and ice water content rises by up to ~120 % in the anvil. SIP also reduces outgoing longwave radiation (OLR) by 6.8 W m–2 (–2.5 %), increases outgoing shortwave radiation (OSR) by 7.9 W m–2 (+2 %), over a 110 km × 110 km region in 6 hours. For this case, precipitation becomes more organized near the convective core, with total accumulation increasing by ~25 %. Under polluted (Nd = 800 cm–3) and clean (Nd = 200 cm–3) conditions, SIP effects are weaker or spatially fragmented, indicating non-linear aerosol dependence. Sensitivity experiments show that, when combined with rime-splintering (RS), Mode 1 (droplet fragmentation during freezing) weakens at high CCN owing to reduced supercooled raindrop supply in the mixed-phase layer, while Mode 2 (drop–ice collisions with splashing) is inhibited at low CCN due to less encounters with large rimed ice particles. In contrast, RS and its combination with ice–ice collisional breakup show relatively limited aerosol sensitivity in this case.
This article investigates the influence of contrasted aerosol conditions through variations in cloud condensation nuclei (CCN) concentrations on secondary ice production (SIP) processes and their impacts on cloud microphysics and convection. The authors perform a series of sensitivity experiments by varying CCN concentrations from clean to polluted conditions for the tropical convective cloud system Hector, using the UK Met Office Unified Model coupled with the CASIM microphysics scheme.
The purpose of this study is highly relevant, as only a limited number of studies have investigated the interaction between aerosols and SIP processes. In most previous works, the effects of aerosols and SIP mechanisms are examined separately. The numerical framework, as well as the range of sensitivity experiments and the chosen diagnostics, appear appropriate to address the scientific questions.
However, several aspects of the manuscript could be improved, particularly the clarity and organization of the results. Given the large number of sensitivity experiments, the presentation of the results is sometimes a bit difficult to follow. For example, some figure panels are discussed long after others from the same figure, which complicates the reading. Further efforts to improve the structure and layout of the figures would also greatly improve the readability of the manuscript.
From a scientific perspective, I also note that the study primarily investigates the sensitivity to CCN concentrations rather than to aerosol conditions in a broader sense. This is because the heterogeneous ice nucleation parameterizations used (Cooper, 1986 and Bigg, 1953) do not explicitly account for aerosol number or composition and therefore do not represent polluted environments in terms of INPs. As a result, the interpretation of the simulations as representing “contrasted aerosol conditions” should be treated with caution. Other aspects of the results would also benefit from additional discussion and clarification.
Overall, the manuscript is suitable for publication in ACP, as the scientific concept of the study is relevant. I therefore recommend publication after minor revisions, which should focus on improving the organization and clarity of the results, as well as on providing a clearer discussion of key physical interpretations and some caveats.
Comments:
Lines 30–32: Additional references could be added here to support the statements.
Lines 40–42: The original work of Phillips et al. (2018), which first introduced the partition of drop fragmentation into two modes, could be cited here.
Line 47: The reference to Phillips et al. (2017) is not fully appropriate here, as this publication focuses on the formulation of the ice breakup parameterization. Part II of the study, which presents a modeling application of the new parameterization, would be more suitable (https://doi.org/10.1175/JAS-D-16-0223.1) as it depict its effect on supercooled liquid water.
Lines 66–67: The discussion focuses exclusively on CCN concentrations. However, more polluted conditions may also imply higher INP concentrations, which could lead to more glaciated clouds with less supercooled liquid water. I therefore suggest restricting this statement explicitly to CCN effects, as stated later in lines 68–71.
Lines 92–95: Is the subgrid cloud fraction scheme still applied in the present study, despite the 1.5 km horizontal resolution? While convection is not parameterized, is it still relevant to parameterize cloud fraction and liquid condensation? Is the liquid fraction not directly predicted by CASIM?
Lines 109–110: Since heterogeneous ice nucleation in the model depends only on temperature (Cooper, 1986) and droplet volume (Bigg, 1953), and not on aerosol number or composition, the sensitivity experiments explore variations in CCN concentration only. In reality, contrasted clean and polluted aerosol environments are expected to modify both CCN and INP concentrations (and properties), with a direct impacts on the ice phase through heterogeneous ice nucleation. This coupled CCN-INP effect is not represented in the present framework and therefore constitutes an important limitation that should be clearly acknowledged. As a consequence, references to clean or polluted environments throughout the manuscript should be explicitly restricted to CCN conditions. The title of the paper might also be reconsidered to refer to CCN conditions rather than aerosol conditions more generally.
Line 172: Why is the RS process considered together with M1, M2, and BR, rather than analyzing these SIP processes individually? For instance, interpreting the effect of RSM1 alone may be challenging, as RS effects could dominate and obscure the specific contribution of M1.
Line 178: How is radar reflectivity calculated in the model? A brief description of the radar operator would be helpful here.
Lines 212–213: While the presence and location of the cloud system are captured, it is evident that the minimum OLR values are higher in all simulations compared to observations. This discrepancy is important and should be at least mentioned.
Section 3.1 : The experiment names could be used directly in the text instead of repeatedly describing each configuration. This would significantly improve readability, as the experiment names already depict both the CCN concentration and whether SIP is active.
Overall, the order in which figures and panels are discussed should be made consistent with their layout. For example, in line 224, Figure 3 is introduced starting with panel (b) but it would be clearer to begin with panel (a).
Section 3.1 is overall descriptive and the physical explanations are provided later in Section 3.3. It would be helpful to inform the readers earlier that the underlying mechanisms are discussed in an upcoming section. The same comment applies to Section 3.2 regarding precipitation.
Figures 3 and 5: Panels corresponding to SIP and no-SIP simulations could be arranged side by side, with consistent sizes, to facilitate direct comparison. Figure 8 provides a good example of a clear and effective layout.
Figure 7: The axis labels and legends are difficult to read. I suggest enlarging this figure, for example by using a two-column format with five rows.
Line 335: Panel 7c is mentioned before panel 7b, which focuses on raindrop number and is not discussed here. The order should be revised.
Lines 340–341: This result is somewhat surprising, as SIP typically generates small ice particles. Previous studies (e.g., Dedekind et al., 2021) have shown that SIP tends to produce smaller ice particles with reduced sedimentation. This could be mentioned.
Lines 343–344: SIP increases ice particle number by several orders of magnitude, while ice water mass increases by only a factor of about two. Given this discrepancy, how can the mean particle diameter increase?
Lines 353–356: While enhanced vapor deposition for ice particles has been reported in previous studies (e.g., Dedekind et al., 2021), why is this effect apparent for riming but not for vapor deposition here? Why is the impact stronger on graupel than on snow? If ice mixing ratio and particle size increase, why does snow not respond similarly?
Lines 363–364: What about the role of water vapor availability for deposition?
Line 393: Figure 7b is referenced here, even though Figure 7 began to be discussed much earlier. This is a bit confusing for the readers. The section can be reorganized to group all discussions of Figure 7 together.
Lines 404–406: This explanation is a bit difficult to follow. Referring directly to the experiment names, rather than to “Nd = …” and “SIP/noSIP”, would improve the clarity of this sentence. This approach could also be applied elsewhere in the manuscript (e.g., lines 451–453).
Line 410: Can you specify which figure is discussed ?
Line 423: When multiple SIP processes are active (e.g., RSM1), it is difficult to disentangle the contribution of each process. An alternative approach could be to analyze differences such as RS–RSM1, but I leave this up to the authors consideration.
Section 4: This section provides a clear and useful summary of the results, clarifying the main findings presented in the previous sections.
Lines 526–527: These results should be balanced with other studies that have shown SIP can reduce particle size and precipitation efficiency (e.g., Dedekind et al., 2021).
Lines 560–561: This statement is not universally valid and may hold only for this specific case, which is particularly important to state given that the following sentence mention a study reporting opposite effects.
Lines 588–589: I do not see a clear link between the cited papers and the purpose of this sentence.
Lines 628–631: In my view, this suggests that snow amounts might also increase, as higher ice number, size, and mass should favor conversion to snow.
Lines 638–639: It should be noted that the simulations still differ significantly from observations.
Conclusions: In addition to uncertainties related to the microphysical assumptions of the model, it should be emphasized that the results strongly depend on the SIP parameterizations used. The physical understanding of SIP mechanisms and therefore their parameterizations remains subject to large uncertainties and ongoing debate. This important limitation should be clearly acknowledged in the conclusions.