Variations of LWC, cloud droplet concentrations, N100, N500 and the 2D images obtained during the first cloud penetration, are shown in Fig. 4. The radar cross section along the flight track at 0220 UTC is shown in Fig. 5. It is shown that, in the lower and higher parts of the supercooled cloud, where the ambient temperature varied from -1°C to -2.6°C and from -4°C to -7°C, the N100 and N500 measured by CIP were generally less than 50 L-1 and 5 L-1, respectively. The high ice number concentrations region, identified by N100, exceeded 100 L-1 and was mainly obtained during 0220 to 0225 UTC, with ambient temperatures ranging from -2.6°C to -3°C. In this region, N100 and N500 apparently increased by a factor of five to ten compared with those in other regions of the penetration, and their maximum values exceeded 300 L-1 and 30 L-1, respectively. Meanwhile, cloud droplet concentrations measured simultaneously by the CAS probe during the first penetration confirmed that this region was a mixed-phase cloud, due to the appearance of a large number of cloud droplets. As shown in Fig. 4a, throughout the majority of the first penetration, cloud droplet concentrations were greater than 20 cm-3, and the maximum values of 80-120 cm-3 appeared in the high ice number concentration area. The LWC derived by the CAS probe was usually lower than 0.01 g m-3 in the lower and higher parts of the cloud, except for the high ice number concentration area. The LWC in the high ice number concentration area exceeded 0.015 g m-3, and the maximum value was approximately 0.025 g m-3. These results suggested that larger cloud droplets may have existed in this region.
The HM mechanism, as investigated in the laboratory and in situ observations of (Mossop, 1985), shows a dependence on the coexistence of drops with diameter <12 μm and >25 μm, with a lesser effect of small drops. Therefore, we also present the number concentrations of large modes of the droplets obtained during the first penetration (Fig. 4b). The variations of large cloud droplet (diameter >25 μm) concentrations along the flight track were consistent with those of N100 and N500, and the maximum value of 0.4 cm-3 was comparable with the value of N100 in the same region. Moreover, the radar reflectivity cross section observed at 0220 UTC demonstrated that the altitudes of cloud top varied from 5-7 km, and cloud top temperatures ranged from -7°C to -18°C along the flight track. The radar reflectivity also indicated that the high ice number concentration region was located in the supercooled part of a shallow convective area, which was approximately 20 km in width and roughly 70-100 km northeast of the radar site. Meanwhile, the maximum reflectivity reached 40 dBZ at 2-3 km in altitude, where the ambient temperature was higher than 0°C (below the melting layer, Fig. 5). According to the radar reflectivities obtained by the C-band radar from 0210 to 0230 UTC, this convective cloud was in a steady state during the first penetration.
Figure 6 illustrates the typical ice particle types obtained by CIP at several points during the first cloud penetration (a to g points as indicated in Fig. 4). The columnar and needle-like ice crystals, typically approximately 300-500 μm in length, were predominant in the high ice concentration area, as indicated by the 2D images. The ice particle types in the high ice concentration region were similar to those reported in previous studies (Rangno and Hobbs, 2001; Crawford et al., 2012) in middle level stratus clouds in South England and Arctic stratus clouds. Meanwhile, aggregates, and some moderately and lightly rimed ice particles (or graupels) with diameters from 500-1000 μm, were also obtained in the same region. It should be noted that columnar and needle-like ice crystals were measured in the southwest of the main convective area (Figs. 6a, b and c), where the ambient temperatures ranged from -0.5°C to -2.25°C. However, their maximum lengths of 500-1000 μm were usually longer than their counterparts at Figs. 6d and e. The needles occurred in the southwest part of the main convective cloud, most likely due to the divergence of air at the cloud top. Then, these needles fell from higher altitudes and kept growing via deposition in the surrounding stratiform cloud. Note that many spherical particles with diameter >100 μm, which may be freezing drops, also appeared in some regions (as indicated by the 2D images at a, b and c points illustrated in Figs. 6a-d). In the higher parts of the cloud, from -3.5°C to -7.5°C, the ice particle types were much more uniform compared with t hose mentioned above. Most of the ice particle types appeared to be heavily and lightly rimed stellar ice crystals (Figs. 6f and g), which we would expect to be produced at temperatures of around -15°C, while some spherical particles were also obtained in this region.
Vertical profiles of ice particle number concentrations from CIP obtained during the first penetration for flight legs in different temperature regions are shown in Fig. 7. The plot, which shows the variations of the 5th, 25th, 50th and 95th percentiles of N100 and N500, indicates that N100 and N500 were highly variable at -3.2°C, where the HM process may have been active, as compared with other parts of the cloud penetration. The median value of N100 (30 L-1) during the first penetration was comparable with previous in situ observations. For example, (Crosier et al., 2011) showed that the median value of irregular ice particles, measured within a weakly convective cloud embedded in a supercooled mid-level stratus cloud in England, was approximately 5 L-1. (Crawford et al., 2012) demonstrated that ice number concentrations, obtained in a mature convective cloud, was approximately 30 L-1 and the maximum value exceeded 100 L-1.
Combined particle size spectra (10 s averaged) measured simultaneously by CAS, CIP and PIP at different points are shown in Fig. 8. The size spectra of large particles were flatter within the high ice number concentration region compared with those in the surrounding area in lower and higher parts of the cloud. Figure 8 also shows that the ratio of size spectra between Fig. 8c and e and between Figs. 8a and b were maximized for diameters from 500 μm to 1000 μm. Additionally, the cloud droplet spectra in the high ice concentration region (or in the higher part of the shallow convective region, Fig. 8) showed that the larger cloud drops (with diameters from 10 μm to 50 μm) increased by a factor of one to five compared with those in other parts of the penetrations.
In this study, N100 and N500 were related to both LWC (Fig. 9) and the large cloud droplets (diameter >24 μm) derived by the CAS probe. This finding was similar to that of (Heymsfield and Willis, 2014), who also presented a relationship between ice particles with diameter >125 μm and large cloud droplets measured in tropical maritime convective clouds during the African Monsoon Multidisciplinary Analyses (NAMMA). However, there are still some discrepancies. For example, although the cloud droplet concentrations were comparable, the LWCs (usually less than 0.015 g m-3) and number concentrations of the large cloud droplets derived by the CAS probe were much smaller than those presented in (Heymsfield and Willis, 2014) within the high ice number concentration zone in a marine convective cloud. These results implied that the large cloud droplet number concentrations were smaller than those measured in NAMMA. In fact, the LWC derived by the CAS probe in the high ice number concentration zone in this study was significantly smaller than that reported by most previous studies, especially those measured in cumulus clouds with intense updraft. For example, (Crawford et al., 2012) demonstrated that the maximum value of LWC in aged wintertime cumulus clouds over the United Kingdom usually exceeded 0.5 g m-3. (Mossop, 1978) found that a large droplet concentration of 10 cm-3 is necessary for an efficient operation of the HM process in cumulus. (Rangno and Hobbs, 2001) also found that the maximum value of LWC in an Arctic stratocumulus cloud exceeded 0.2 g m-3.
On the other hand, large cloud droplets and high values of LWC are not always captured in some clouds where ice multiplication apparently appeared. Observations conducted recently also presented concurrences of low LWC and high ice number concentration in weakly convective cloud and maritime chimney cloud. For instance, based on data gathered in England, (Crosier et al., 2011) demonstrated that LWC and large cloud droplets (diameter >10 μm), within the temperature range -4°C to -5°C, were usually less than 0.04 g m-3 and 0.1 cm-3, respectively, while the ice number concentrations were usually greater than 30 L-1 and the types of ice particles were predominated by pristine needles and columns. Additionally, based on data obtained in chimney cloud during the ICE-T project, (Heymsfield and Willis, 2014) concluded that secondary ice particles were observed primarily in regions of low LWC and weak vertical velocity, while LWC in the regions where secondary ice crystals were observed were dominantly below 0.1 g m-3, with a median value of only 0.03 g m-3. These data are comparable with those obtained in the present study in the same temperature range. We speculate that the updraft in this case was relatively weaker compared to that in typical convective cloud. Therefore, the majority of droplets may have been removed quickly by the Bergeron-Findeisen process after high ice concentrations were produced via ice multiplication, which resulted in the coexistence of large droplets and ice crystals for a few minutes only. Thus, it is hard to capture large droplets and high ice concentrations simultaneously.
During 0240-0300 UTC, to obtain the vertical microphysical properties and compare them with those in the first penetration, a quasi-Lagrange (fast descent in 10 minutes) sampling was performed within the cloud north of the first penetration. Unfortunately, 3 minutes of data (from 0243 to 0246 UTC) with temperatures ranging from -4°C to -3°C were not uploaded to the data collection system due to hardware problems. Radar data in the same region were also unavailable because its wave band was blocked by buildings (Fig. 3b). Therefore, it was hard to estimate whether this area was a convective or a stratiform cloud and whether the high ice number concentrations appeared in this region. Despite this inherent shortcoming, the rest of the data showed (Fig. 10) that concentrations of N100, N500, LWC and cloud droplets were much smaller than their counterparts obtained during the first penetration. For example, N100 and N500 were usually less than 8 cm-3 and 3 cm-3, respectively, while the LWC and cloud droplet concentrations derived from the CAS probe (1-50 μm) were usually smaller than 0.01 g m-3 and 60 cm-3, respectively. The maximum values mainly occurred in the higher and lower parts of the cloud, with temperature ranging from -5°C to -6°C and -0.3°C to -1°C, respectively. The concentrations of large cloud droplets decreased by a factor of three to five compared with those in the high ice concentration area in the first penetration. Meanwhile, the 2D images obtained by CIP showed large discrepancies compared with those in the first penetration (Fig. 11). Sector plates and rim ice particles, with diameter >1000 μm, occurred in the higher part of the cloud, where ambient temperatures were lower than -5°C. The heavily rimed ice particles dominated in the lower part of the cloud, where the ambient temperatures were higher than -3°C. The occurrence of rimed ice particles could be attributed to the riming growth of ice particles during free fall. The pristine ice crystals, such as column and needle crystals, which may relate to secondary ice production, were very rare throughout the second penetration (even just below the missing data area), and freezing drops were also absent. It was possible that the updraft during the second penetration was relatively small because the cloud was in its dispersion stage. Therefore, the majority of the supercooled cloud droplets and ice particles were removed by evaporation/sublimation, leading to only very low concentrations of large cloud droplets and ice particles, which were insufficient for the activation of the HM process. Overall, it was unlikely that secondary ice production was active during the second penetration, and the ice formed in this region may have been primarily from ice nucleation.