Efficient ice multiplication from freezing raindrop fragmentation

Supplementary information for the article ’Efficient ice multiplication from freezing raindrop fragmentation’ Nils Pfeifer1*, Bernd Mom2, Dmitri Moisseev2,3, Susan Hartmann4, Julian Meusel5, Corinna Hoose5, Maximilian Maahn1 1*Leipzig Institute for Meterology (LIM), Leipzig University, Leipzig, Germany. 2Institute for Atmospheric and Earth System Research/Physics, University of Helsinki, Helsinki, Finland. 3Finnish Meteorological Institute, Helsinki, Finland. 4Department of Atmospheric Microphysics, Leibniz Institute for Tropospheric Research, Leipzig Germany. 5Institute of Meteorology and Climate Research Troposphere Research, Karlsruhe Institute of Technology, Karlsruhe Germany. 1 Supplementary Discussion 1.1 Melting and refreezing time From the layer of reduced linear depolarisation ratio (LDR) (Figure 1.a) below the melting zone, it can be deduced that the snowflakes have collapsed into a spherical shape. This can be confirmed visually by the emergence of drops in Figure 1.c, starting from 14:30 UTC. The later observed ice pellets can be produced from these drops in three ways. Either the drops are completely liquid and start freezing after supercooling and a nucleation event, they freeze after contact with an ice splinter, or they retain some ice and immediately start refreezing upon exiting the melting layer [1]. Since the temperature (Fig. 2.a) is high with respect to ice nucleating particle (INP) activity below the melting layer, the influence of nucleation is regarded as minor. We can not determine to which fraction ice is still contained in the raindrop after exiting the warm layer as the microphysics of snowflakes is not well understood [2] and the original geometry of the snowflakes melting into drops is unknown. In order to obtain 1 an approximate estimation of the size up to which snowflakes are undergoing melting, the model for falling drops outlined in [3] is employed, with the specific case of ’melting of a solid ice particle while falling’. This is motivated by the observation that before the transition to rain and ice pellets most snowflakes observed at the ground have a compact, roundish shape and assuming that the snowflakes falling into the melting layer are similar. The exception here are the columns, which are expected to melt completely before entering the sub 0 °C layer. The maximum (+1 °C) and minimum (+0.1 °C) values of the warm layer from the sounding profile in Figure 2.a are used. For the sedimentation velocities, 2 m/s and 1 m/s are considered as typical values measured at the ground by the Video In Situ Snowfall Sensor (VISSS) for drops. A fraction of drops actually falls faster than this (up to 5 m/s), but it can be assumed that the initial velocity of snowflakes entering the warm layer is much lower. Therefore, it can be argued that the chosen values of the velocity are valid from a statistical perspective. The thickness of the melting layer is assumed to be 540 m. Regarding a fall velocity of 2 m/s (1 m/s), for the case of an environmental temperature of +0.1 °C, frozen drops would melt up to a size of approximately 250 µm (400 µm) within 4.5 min (9 min). For the maximum temperature of +1 °C, hydrometeors up to a diameter of 900 µm (1250 µm) would melt completely. However, according to the vertical temperature profile only half of the time the temperature is significantly above the minimum value. Therefore, the size limit for complete melting is closer to the values for +0.1 °C. Despite the uncertainties in geometry and velocity of the initial snowflakes, it can be deduced that at least some of the drops still contain ice and start refreezing immediately after leaving the melting layer. This can result in drop fragmentation upon freezing (DFF), for which the complete freezing of the drops is not necessary. The growth of an ice shell around the drops is sufficient [4,5]. Whether this ice shell growth also occurs in the case of ice remaining in the drop is currently unknown. The fragmentation can lead to the production of ice splinters, which in turn can serve as catalysts for further freezing, thereby initiating a cascading effect. Nevertheless, the relative contributions of these processes to the glaciation of precipitation remain uncertain. 1.2 Analysis of radar Doppler spectra Figure 1displays radar Doppler spectra of the LDR captured by the vertically pointing cloud radar during the different phases from the lowest range gate (102.2 m) up to a height of 1000 m. The points in time when these spectra were measured are indicated in Fig. 1.a by white dashed lines. The distribution of hydrometeors during phase I above 300 m height is localized in Doppler velocity bins between 0 and 4 m/s with values above -20 dB, which can be interpreted as snowflake aggregates. A maximum with LDR values reaching up to -15 dB close to 0 m/s is visible. This is suggesting the presence of column monomers, since these are associated with high values of LDR and slow sedimentation velocities [6]. Disturbances in the distribution due to turbulence become apparent below 300 m. In Phase II, the majority of the spectrum exhibits low values of LDR close to to the minimal observable value of -30 dB, indicating a collapse into spherical shapes due to melting for most of the velocity bins and heights. However, a single mode of increased LDR for the highest velocities (appr. 8 - 9 m/s) below 500 m altitude is visible and indicates refreezing of at least some of the largest 2 drops into ice pellets. Asymetric ice shell growth and particle wobbling can lead to an enhancement in LDR [7] that can be seen in the refreezing signature in ??.a during Phase II. In contrast to this single mode, the Phase III is characterized by a bimodal LDR spectrum. A second mode of hydrometeors, which commences at approximately 15:10 at an altitude of 700 meters and reaches the ground at approximately 15:25, exhibits velocities just above 0 meters per second. Following the criteria given in [6], this slow mode can be again classified as columnar ice. The maximum altitudes of this mode are observed at levels that are marginally lower than the maximum altitudes of the first, fast falling mode. This suggests that the fast mode may be caused by the fraction of larger drops that have refrozen, while the slower mode can be attributed to columnar ice growing from small splinters that are likely ejected during DFF secondary ice production (SIP). These splinters require time to transform into an elongated form, which is discernible in the LDR spectra. The second mode reaches the ground at 15:25 and vanishes at 17:00, overlapping with the increased number concentration. Eventually, in phase IV, the disappearance of the fast mode is observed as well. This indicates the completion of the transition to rain. Supplementary Figure 1: Height resolved radar Doppler LDR spectra from selected temporal points during the four phases of the event. These spectra were observed by the 94 GHz cloud radar. a) Phase I: Columnar Ice and aggregates falling from upper layers before the beginning of melting. b) Phase II: Raindrops resulting from melting an cluster of irregularly shaped particles resulting from refreezing at high velocities. c) Phase III: Fast mode of irregularly shaped particles resulting from refreezing, slow mode of columnar ice, resulting from SIP.d) Phase IV: Rain. Acronyms VISSS Video In Situ Snowfall Sensor SIP secondary ice production INP ice nucleating particle DFF drop fragmentation upon freezing LDR linear depolarisation ratio 3 Supplementary References [1] Tobin, D.M., Kumjian, M.R., Oue, M., Kollias, P.: Refreezing of Partially Melted Hydrometeors: Polarimetric Radar Observations and Microphysical Model Simulations. Journal of the Atmospheric Sciences 80(3), 725–741 (2023) https://doi. org/10.1175/JAS-D-22-0174.1 . Accessed 2024-05-06 [2] Leinonen, J., Lerber, A.: Snowflake Melting Simulation Using Smoothed Particle Hydrodynamics. Journal of Geophysical Research: Atmospheres 123(3), 1811–1825 (2018) https://doi.org/10.1002/2017JD027909 . eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/2017JD027909. Accessed 202504-22 [3] Mason, B.J.: On the melting of hailstones. Quarterly Journal of the Royal Meteorological Society 82(352), 209–216 (1956) https://doi.org/10.1002/qj.49708235207 . eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/qj.49708235207. Accessed 2025-02-11 [4] Keinert, A., Spannagel, D., Leisner, T., Kiselev, A.: Secondary Ice Production upon Freezing of Freely Falling Drizzle Droplets. Journal of the Atmospheric Sciences 77(8), 2959–2967 (2020) https://doi.org/10.1175/JAS-D-20-0081.1 . Accessed 2024-05-06 [5] Kleinheins, J., Kiselev, A., Keinert, A., Kind, M., Leisner, T.: Thermal Imaging of Freezing Drizzle Droplets: Pressure Release Events as a Source of Secondary Ice Particles. Journal of the Atmospheric Sciences 78(5), 1703–1713 (2021) https: //doi.org/10.1175/JAS-D-20-0323.1 . Publisher: American Meteorological Society Section: Journal of the Atmospheric Sciences. Accessed 2025-04-07 [6] Li, H., M¨ohler, O., Pet¨aj¨a, T., Moisseev, D.: Two-year statistics of columnar-ice production in stratiform clouds over Hyyti¨al¨a, Finland: environmental conditions and the relevance to secondary ice production. Atmospheric Chemistry and Physics 21(19), 14671–14686 (2021) https://doi.org/10.5194/acp-21-14671-2021 . Publisher: Copernicus GmbH. Accessed 2025-02-04 [7] Tobin, D.M., Kumjian, M.R.: Microphysical and Polarimetric Radar Modeling of Hydrometeor Refreezing. Journal of the Atmospheric Sciences 78(6), 1965–1981 (2021) https://doi.org/10.1175/JAS-D-20-0314.1 . Publisher: American Meteorological Society Section: Journal of the Atmospheric Sciences. Accessed 2024-12-27 4