Review of “Modeling the Coupled and Decoupled states of Polar Boundary-Layer Mixed-Phase Clouds” by Étienne Vignon, Lea Raillard, Audran Borella, Gwendal Rivière, and Jean-Baptiste Madeleine. Submitted for publication in EGUsphere October 6, 2025.

This paper evaluates two new microphysical parameterizations in simulations of well-tested M-PACE and ISDAC mixed-phase stratocumulus cases in the LMDZ (global atmospheric component of the IPSL-CM Earth System Model) single column model. In this model, boundary layer turbulent fluxes are parameterized with an Eddy Diffusivity-Mass Flux scheme, where the mass-flux scheme is only active when surface convective instability occurs. Therefore, turbulence in decoupled cloud cases (i.e., the ISDAC case) is only parameterized with local counter-gradient diffusion.   

In the current version of the model, phase-partitioning in boundary layer clouds is a function of temperature. A parameterization developed in Raillard et al. (2025) for mid-level clouds that replaces a temperature dependent formulation for one that is a function of subgrid turbulent activity and ice crystal properties is added to the convective boundary layer scheme. The second new parameterization adds a “homogenization” term to the equation for the evolution of supersaturation of ice. This parameterization accounts for air parcels mixing between clouds in the environment and air in the surface-forced thermal plumes. This parameterization was included in Furtado et al. (2016) but not in Raillard et al. (2025). This second parameterization is only active when surface convective instability occurs. Simulations without these new parameterizations is referred to as CNTL. Simulations with the new phase-partitioning scheme is referred to as R25. Simulations with both new parameterizations is referred to as TEST.

Perturbed parameters ensemble experiments are performed for the two case studies to test the sensitivity to parameters that control turbulence and ice crystal properties within acceptable ranges.

The main results of this study are:

For the M-PACE case, only the TEST simulation can produce a mixed-phase stratocumulus with cloud liquid and ice similar to the observations. Even though the R25 simulation has a more realistic potential temperature profile, it produces almost no liquid and too much cloud ice.

Questions:

1. The fixed Naero5 for the M-PACE case should be much lower than for the ISDAC case. 0.16/L is the value typically used for these cases studies. How does R25 perform with lower values of INP?
2. TEST produces relative humidity with respect to liquid in the liquid layer that is greater than 100%. How is this possible?
3. TEST has near surface layer RH that is much lower than obs and the other runs, indicating too much mixing of sub-cloud air into the liquid layer?
4. M-PACE has large surface fluxes (cold air outbreak) but is there no representation of cloud-driven top-down turbulence in the model? How does this change the balance between turbulence and microphysics at cloud top in the model?
5. TEST produces large TKE above the liquid layer in the inversion (Figure 3c). How is TKE calculated?
6. Lines 341-345: How is turbulence due to cloud-top radiative cooling represented in the model? By local diffusion? Or is it in the microphysical scheme, determined by the Lagrangian turbulent decorrelation time-scale? Is it assumed that local diffusion estimated from TKE is a representation of non-local mixing by cloud-top cooling? Is so, why is TKE limited to cloud top in the ISDAC simulation?

For the ISDAC case, since this is a decoupled case where the surface convective scheme is inactive, R25 and TEST are the same. This simulation tests the impact of the two different phase-partitioning schemes.

Questions/Comments:

1. It is no surprise that the temperature dependent phase-partitioning scheme produces too much cloud ice and causes the liquid layer to collapse.
2. Lines 367-369: Again top-down vertical diffusion of TKE by subgrid turbulence is discussed but isn’t this just local diffusion in the model?
3. Figure 6c shows TKE buoyancy term and TKE essentially only at cloud top but Ovchinnikov et al. (2014) shows maximum TKE near the liquid cloud base. Even though the R25/TEST simulations maintain a liquid layer, this result indicates fundamental error in turbulence that will also produce fundamental errors in microphysics.

Minor comments:

• Line 98: Change “ofrid” to “of”.
• Figure 2 and 3: Include “M-PACE” in figure caption.
• Line 343: “…loop, that…”

Summary:

I have major questions about the parameterizations used in this climate model.

• Is it assumed that local diffusion estimated from TKE is a representation of non-local mixing by cloud-top cooling? If so, why is TKE limited to cloud top in the ISDAC simulation?
• Also, There are many ways to modify turbulence/microphysics in order to maintain cloud liquid. My concern is that this simulation is getting the right answer for the wrong reason. This is extremely important for climate simulations since it will produce unrealistic sensitivity to changes in environmental conditions, surface conditions, and aerosols.

Overview

The authors present a clearly written model development work that relies upon single-column model (SCM) simulations of two well-studied cases of mixed-phase cloud (MPCs). Their modeling approach takes on the revision of phase partitioning to account for the production of supercooled liquid water in updraft regions of turbulent layers that may or may not be coupled to the surface. Another approach that has also proven effective (e.g., Silber et al. 2022 appendix) is application of microphysical process rates in a procedure where a moist turbulence scheme operates on thermodynamic and microphysics fields (e.g., mixing ice species), cloud liquid water is diagnosed from a macrophysics scheme (rapid equilibration), ice formation rate is diagnosed from ambient aerosol-modulated immersion freezing (in addition to multiplication schemes), ice cover is diagnosed by macrophysics, and ice growth then follows from thermodynamic conditions with associated sedimentation offset by turbulent mixing. In the latter approach, phase partitioning is only indirectly affected by turbulence. It would be very interesting to see this new parameterization approach tested against other approaches in a case with rapidly evolving boundary layer depth and cloud-top temperature, such as the ongoing COMBLE-MIP community project (https://arm-development.github.io/comble-mip/README.html). I see no major methodological flaws in the presented work and recommend publication of this relevant work after addressing my comments below (especially comment 3).

Comments

1. Regarding the second paragraph of the introduction, I would urge the authors to consider another much simpler approach to explaining polar mixed-phase clouds (e.g., Silber et al., 2021), namely by considering their features as a readily understood outcome of weak heterogeneous ice formation from their familiar warm (liquid-only) counterparts. This allows a more intuitive explanation of many widely observed key features. For instance, in a well-mixed coupled case in the limit of weak ice formation (approaching the warm cloud case), cloud liquid is roughly adiabatic (consistent with meteorology 101). Once heterogeneous ice formation is sustained in the immersion mode at the very weak levels it is typically observed (cf. Silber et al., 2021), it serves as a very weak sink of moisture and naturally "the liquid remains at the top". This also readily explains the resilience, which I think should not be unexpected at all and is reproduced by higher resolution models. I therefore suggest to avoid furthering the "unexpected" idea by repeating it here, because it is based on the mistaken baseline assumption that there is no relevant time scale to ice growth and sedimentation. Taking the warm case as a foundation also readily explains the commonality of continuous liquid bases observed by lidar (see examples in Silber et al., 2021), which becomes circuitous in the current explanation owing to dependence on updrafts (note: adiabatic liquid water content in a well-mixed layer is independent of updraft strength). In my opinion, referring to WBF also introduces an unnecessary overlay that is not encoded in models as a "process" because models don't need to add anything to the basic physics: namely, ice is growing everywhere that relative humidity exceeds saturation with respect to ice, continuously both above and below any supercooled liquid cloud bases and there is no need to consider any further explanation separately within versus below liquid cloud base.

2. Regarding the introduction to model capabilities (lines 60), I would suggest to add more than one reference showing that many higher resolution models can perform very well indeed for both coupled and decoupled MPCs as long as ice formation rate and ice properties are realistic. For instance, for a coupled case, Tornow et al. (2025) illustrate mixed-phase simulations that can reproduce basic features of sustained liquid water path, precipitation onset and subsequent cloud cover and droplet number concentration reduction. For a decoupled case, Silber et al. (2019, 2020) present large-eddy simulations that reproduce the progressive development of supercooled liquid in a stable layer, turbulence onset, and in that case, the sustained coexistence of liquid and ice precipitation processes. I would also add that the Lagrangian approach taken in those studies provides a strengthened foundation for large-scale model development because it allows a test of whether a mixed-phase cloud can realistically form within an initially cloud-free environment and proceed to reproduce observed transitions. Silber et al. (2022) also illustrate that the NASA ModelE3 GCM code can well reproduce the decoupled cloud case in single-column model (SCM) mode (see appendix), including onset of turbulence and co-existing precipitation in two phases.

3. Regarding the leading problems in large-scale model parameterization, I would have placed first the extreme uncertainty in parameterization of ice formation rate in the immersion mode (e.g., Knopf et al., 2023) and via ice multiplication where it far outpaces the immersion mode (e.g., Korolev et al., 2024; likely same process as in deeper convection from long evidence of colocation with drizzle in MPCs). These are but a few examples of decades of evidence that we cannot capture ice formation rates to one or even several orders of magnitude. No degree of improving other processes can readily cover for that. The Knopf et al. (2023) study further shows how a diagnostic scheme following DeMott type INP parameterizations can produce extremely unrealistic rates of ice formation. Can the authors show that their ice formation rates in these cases are consistent with simple rough estimates of INP source strength and activation rate?

4. Regarding the case studies selected (section 2.2), a major difference is that ISDAC included continuous nudging of temperature and water vapor whereas M-PACE applied fixed large-scale advective flux divergence profiles. When applying nudging to LES and SCM at every time step, the model thermodynamic profiles cannot diverge as much from one another; in other words, if divergence grows more in one model, it is more offset. Please clarify for readers in the text whether the authors apply nudging in the ISDAC case and whether that differs from the M-PACE case.

References

Knopf, D.A., I. Silber, N. Riemer, A.M. Fridlind, and A.S. Ackerman, 2023: A 1D model for nucleation of ice from aerosol particles: An application to a mixed-phase Arctic stratus cloud layer. J. Adv. Model. Earth Syst., 15, no. 10, e2023MS003663, doi:10.1029/2023MS003663.

Korolev, A., Z. Qu, J. Milbrandt, I. Heckman, M. Cholette, M. Wolde, C. Nguyen, G. McFarquhar, P. Lawson, and A. Fridlind, 2024: High ice water content in tropical mesoscale convective systems (a conceptual model). Atmos. Chem. Phys., 24, no. 20, 11849-11881, doi:10.5194/acp-24-11849-2024.

Silber, I., A.M. Fridlind, J. Verlinde, A.S. Ackerman, Y.-S. Chen, D.H. Bromwich, S.-H. Wang, M. Cadeddu, and E.W. Eloranta, 2019: Persistent supercooled drizzle at temperatures below -25°C observed at McMurdo Station, Antarctica. J. Geophys. Res. Atmos., 124, no. 20, 10878-10895, doi:10.1029/2019JD030882.

Silber, I., A.M. Fridlind, J. Verlinde, L.M. Russell, and A.S. Ackerman, 2020: Non-turbulent liquid-bearing polar clouds: Observed frequency of occurrence and simulated sensitivity to gravity waves. Geophys. Res. Lett., 125, no. 10, e2020GL087099, doi:10.1029/2020GL087099.

Silber, I., A.M. Fridlind, J. Verlinde, A.S. Ackerman, G.V. Cesana, and D.A. Knopf, 2021: The prevalence of precipitation from polar supercooled clouds. Atmos. Chem. Phys., 21, no. 5, 3949-3971, doi:10.5194/acp-21-3949-2021.

Silber, I., R.C. Jackson, A.M. Fridlind, A.S. Ackerman, S. Collis, J. Verlinde, and J. Ding, 2022: The Earth Model Column Collaboratory (EMC2) v1.1: An open-source ground-based lidar and radar instrument simulator and subcolumn generator for large-scale models. Geosci. Model Dev., 15, no. 2, 901-927, doi:10.5194/gmd-15-901-2022.

Tornow, F., E. Crosbie, A. Fridlind, A.S. Ackerman, L.D. Ziemba, G. Elsaesser, B. Cairns, D. Painemal, S. Chellappan, P. Zuidema, C. Voigt, S. Kirschler, and A. Sorooshian, 2025: High accumulation mode aerosol concentration and moderate aerosol hygroscopicity limit impacts of recent particle formation on Northwest Atlantic post-frontal clouds. Geophys. Res. Lett., 52, no. 18, e2025GL116020, doi:10.1029/2025GL116020.