Determine why heterogeneous ice formation is so strongly suppressed in vigorous deep convective clouds that copious supercooled liquid persists to the homogeneous freezing threshold near −38 °C despite entrainment of boundary-layer aerosols containing potential ice nuclei, as first documented by the in-situ measurements of Rosenfeld & Woodley (2000); quantify how often such deep supercooled layers occur globally and with what climatic importance; and reduce this process understanding to reliable mixed-phase parameterizations for the approximately −30 to −40 °C regime.

Background

Cloud water can remain liquid far below 0 °C; such supercooled liquid is metastable and converts to ice once ice is nucleated. Two freezing channels matter in the atmosphere. In heterogeneous ice nucleation, a small subset of aerosol particles — ice-nucleating particles (INPs) — catalyze freezing at temperatures between 0 °C and roughly −38 °C, depending on their surface properties. In homogeneous freezing, ice nucleates spontaneously within the liquid itself; for cloud-sized droplets this becomes essentially inevitable near −38 °C, which therefore acts as a practical floor for liquid water in the atmosphere. Deep convective clouds (cumulonimbus) are vigorous, buoyancy-driven clouds rooted in the boundary layer, the lowest 1–2 km of the atmosphere that contains most aerosols, including particles that act as cloud condensation nuclei (CCN, on which cloud droplets form) and potential INPs. Where liquid droplets and ice crystals coexist, a cloud is called mixed-phase; the conversion of liquid to ice, termed glaciation, controls precipitation efficiency, hail growth, and thunderstorm electrification (charge separation in storms requires coexisting supercooled liquid and ice), so the temperature at which a cloud glaciates — its glaciation temperature — is a key observable and a key parameter for cloud models. The maximum liquid water that a rising cloud parcel can attain by condensation if it ascends without mixing with surrounding air is called the adiabatic liquid water content; observed values well below it indicate dilution by entrainment. The rate at which ice particles accrete supercooled droplets (riming) is governed by the collision efficiency between ice crystals and droplets, which is very small when both are small. Ice number can also grow without any new nucleation through secondary ice production — ice-multiplication processes such as splintering during riming — which can make ice concentrations far exceed the INP count.

Before 2000, highly supercooled liquid had been documented mainly in cirrus and in orographic wave clouds (clouds forming in the vertical air motion over mountains), in small amounts (less than 0.15 g m⁻³), attributed to small droplet size and scarce INPs at those altitudes: Sassen (1985) for cirrus and Heymsfield & Miloshevich (1993) for wave clouds. Deep convective clouds were expected to glaciate much warmer because their tops contain much larger droplets and because they ingest boundary-layer air rich in aerosols, many of which can nucleate ice below about −25 °C. Satellite retrievals suggested otherwise. Rosenfeld & Woodley (2000) reported aircraft in-situ measurements in vigorous continental convective storms over Texas in August 1999 (with supporting Argentinean measurements in early 2000) showing that most of the condensed water remains liquid down to −37.5 °C: liquid water contents up to 1.8 g m⁻³ at −37.5 °C with a median droplet volume diameter of about 17 µm — an order of magnitude more supercooled water than previously reported at such temperatures — with an abrupt transition to ice at slightly colder temperatures, indicating homogeneous freezing as the dominant glaciation mechanism. The coexistence of numerous small droplets with few large ice particles implies that riming was inefficient, and hence that heterogeneous ice formation was strongly suppressed despite abundant boundary-layer aerosol. The observation matters because a deep supercooled layer is the water reservoir for hail growth and thunderstorm charging, and because numerical cloud models of that era (and many since) lack reliable mixed-phase microphysics between about −30 and −40 °C, so simulated rainfall, hail, and electrification in vigorous convection depend on poorly constrained parameterizations. The authors explicitly left open why heterogeneous nucleation is so inefficient in these clouds — suggesting, without a direct test, very small collision efficiencies between sub-200 µm ice crystals and sub-40 µm cloud droplets — and how often such deep supercooled layers occur globally and what their broader climatic importance is.

Problem Statement

Establish a validated, quantitative account of why heterogeneous ice formation is strongly suppressed in vigorous deep convective clouds in the approximately −30 to −40 °C regime, such that copious supercooled liquid (of order 1 g m⁻³, roughly half the adiabatic liquid water content, i.e. roughly half the maximum attainable by undiluted condensational growth) persists up to the homogeneous freezing threshold near −38 °C despite entrainment of boundary-layer aerosols containing potential ice nuclei. This requires: (1) determining the relative contributions of CCN-controlled droplet size, ice-nucleating particle availability, secondary ice production, and updraft dynamics, and in particular directly testing the hypothesis of Rosenfeld & Woodley (2000) that the suppression reflects very small collision efficiencies between sub-200 µm ice crystals and sub-40 µm cloud droplets; (2) mapping the temperature and size dependence of the transition from heterogeneous to homogeneous freezing in realistic convective conditions, including turbulence and varying aerosol loads; (3) quantifying the global frequency of such deep supercooled layers in deep convective clouds and their climatic importance, including consequences for precipitation efficiency, hail growth, and cloud electrification; and (4) reducing this process understanding to reliable mixed-phase parameterizations for the −30 to −40 °C regime in numerical cloud and climate models, replacing the poorly constrained treatments that currently govern simulated rainfall, hail, and electrification in vigorous convection. These components serve one objective: a closed mechanistic and quantitative description of the deep-convective glaciation transition identified by the in-situ observations of Rosenfeld & Woodley (2000).

References

  1. Rosenfeld, D. & Woodley, W. L. Deep convective clouds with sustained supercooled liquid water down to −37.5 °C. Nature 405, 440–442 (2000). DOI 10.1038/35013030, https://doi.org/10.1038/35013030
  2. Sassen, K., Liou, K. N., Kinne, S. & Griffin, M. Highly supercooled cirrus cloud water: confirmation and climatic implications. Science 227, 411–413 (1985). DOI 10.1126/science.227.4685.411, https://doi.org/10.1126/science.227.4685.411
  3. Heymsfield, A. J. & Miloshevich, L. M. Homogeneous ice nucleation and supercooled liquid water in orographic wave clouds. Journal of the Atmospheric Sciences 50, 2335–2353 (1993). DOI 10.1175/1520-0469(1993)050<2335:HINASL>2.0.CO;2, https://doi.org/10.1175/1520-0469(1993)050%3C2335:HINASL%3E2.0.CO;2
  4. Khain, A. P., Rosenfeld, D. & Pokrovsky, A. Simulating convective clouds with sustained supercooled liquid water down to −37.5 °C using a spectral microphysics model. Geophysical Research Letters 28, 3887–3890 (2001). DOI 10.1029/2000GL012662, https://doi.org/10.1029/2000GL012662
  5. Khain, A. & Pokrovsky, A. Simulation of Effects of Atmospheric Aerosols on Deep Turbulent Convective Clouds Using a Spectral Microphysics Mixed-Phase Cumulus Cloud Model. Part II: Sensitivity Study. Journal of the Atmospheric Sciences 61, 2983–3001 (2004). DOI 10.1175/jas-3281.1, https://doi.org/10.1175/jas-3281.1
  6. Pinsky, M., Khain, A. & Korolev, A. Analytical Investigation of Glaciation Time in Mixed-Phase Adiabatic Cloud Volumes. Journal of the Atmospheric Sciences 71, 4143–4157 (2014). DOI 10.1175/JAS-D-13-0359.1, https://doi.org/10.1175/JAS-D-13-0359.1
  7. Yuan, T., Martins, J. V., Li, Z. & Remer, L. A. Estimating glaciation temperature of deep convective clouds with remote sensing data. Geophysical Research Letters 37 (2010). DOI 10.1029/2010GL042753, https://doi.org/10.1029/2010GL042753
  8. Rosenfeld, D., Yu, X., Liu, G., Xu, X., Zhu, Y., Yue, Z., Dai, J., Dong, Z., Dong, Y. & Peng, Y. Glaciation temperatures of convective clouds ingesting desert dust, air pollution and smoke from forest fires. Geophysical Research Letters 38 (2011). DOI 10.1029/2011GL049423, https://doi.org/10.1029/2011GL049423
  9. Carro-Calvo, L., Hoose, C., Stengel, M. & Salcedo-Sanz, S. Cloud glaciation temperature estimation from passive remote sensing data with evolutionary computing. Journal of Geophysical Research: Atmospheres 121 (2016). DOI 10.1002/2016JD025552, https://doi.org/10.1002/2016JD025552
  10. Korolev, A., McFarquhar, G., Field, P. R., Franklin, C., Lawson, P., Wang, Z., Williams, E., Abel, S. J., Axisa, D., Borrmann, S., Crosier, J., Fugal, J., Krämer, M., Lohmann, U., Schlenczek, O., Schnaiter, M. & Wendisch, M. Mixed-Phase Clouds: Progress and Challenges. Meteorological Monographs 58, 5.1–5.50 (2017). DOI 10.1175/AMSMONOGRAPHS-D-17-0001.1, https://doi.org/10.1175/AMSMONOGRAPHS-D-17-0001.1
  11. Yang, J., Wang, Z. & Heymsfield, A. On the freezing time of supercooled drops in developing convective clouds over tropical ocean. Atmospheric Research 211, 30–37 (2018). DOI 10.1016/j.atmosres.2018.04.023, https://doi.org/10.1016/j.atmosres.2018.04.023
  12. Takeishi, A. & Storelvmo, T. A Study of Enhanced Heterogeneous Ice Nucleation in Simulated Deep Convective Clouds Observed During DC3. Journal of Geophysical Research: Atmospheres 123 (2018). DOI 10.1029/2018JD028889, https://doi.org/10.1029/2018JD028889
  13. Correia, A. L., Sena, E. T., Silva Dias, M. A. F. & Koren, I. Preconditioning, aerosols, and radiation control the temperature of glaciation in Amazonian clouds. Communications Earth & Environment 2, 168 (2021). DOI 10.1038/s43247-021-00250-3, https://doi.org/10.1038/s43247-021-00250-3
  14. Hofer, S., Hahn, L. C., Shaw, J. K., McGraw, Z. S., Bruno, O., Hellmuth, F., Pietschnig, M., Mostue, I. A., David, R. O., Carlsen, T. & Storelvmo, T. Realistic representation of mixed-phase clouds increases projected climate warming. Communications Earth & Environment 5, 390 (2024). DOI 10.1038/s43247-024-01524-2, https://doi.org/10.1038/s43247-024-01524-2
  15. Munawar, I., Zhu, Y., Wang, M., Rosenfeld, D., Liu, J. & Wang, Y. The dominant role of aerosol's CCN effect in cloud glaciation. npj Climate and Atmospheric Science 8, 121 (2025). DOI 10.1038/s41612-025-00999-w, https://doi.org/10.1038/s41612-025-00999-w
  16. Ross, T. I. D. & Lasher-Trapp, S. Investigating the Relative Roles of INPs and CCN in a Simulated Thunderstorm Using a New Immersion Freezing Algorithm. Journal of Geophysical Research: Atmospheres 130, e2024JD042592 (2025). DOI 10.1029/2024JD042592, https://doi.org/10.1029/2024JD042592
  17. Huang, S., Yang, J., Li, J., Chen, Q., Zhang, Q. & Guo, F. Impact of secondary ice production on thunderstorm electrification under different aerosol conditions. Atmospheric Chemistry and Physics 25, 1831–1850 (2025). DOI 10.5194/acp-25-1831-2025, https://doi.org/10.5194/acp-25-1831-2025