Riming electrification is the main charge separation mechanism of thunderstorms, occurring mainly during graupel particle–ice crystal collisions. Laboratory experiments have found that charge separation polarity and magnitude depend critically on cloud water content and temperature. Several groups have mapped this dependence, but there are substantial differences between their results. These conflicting laboratory-derived riming electrification topographies can be tested by comparing them to field observations. Here, direct and simultaneous sonde-based measurement of both precipitation particle type and charge (videosonde) and cloud water content [hydrometeor videosonde (HYVIS)] in lightning-active Hokuriku winter clouds at Kashiwazaki, Niigata Prefecture, Japan, are reported. With decreasing height, summed graupel charge transitioned from negative to positive at a mean temperature of −11°C, and the mean peak cloud water content in the positive graupel domain was 0.4 g m−3. Thus, in cloud regions of relatively high temperature (≥−11°C) and low cloud water content (CWC; ≤0.4 g m−3), graupel particles were mainly positively charged. This result can be compared with those of laboratory riming experiments; for example, in this temperature/cloud water content domain, graupel electrification has been reported to be positive by Takahashi, largely negative in early reports using the Manchester cloud chamber, positive in later reports using the Cordoba and Manchester modified cloud chambers, and partially positive in a more recent report using the Cordoba cloud chamber.
Clouds are electrified by precipitation particle interaction. Field observations have confirmed that graupel particles are critically important in this process (Dye et al. 1986; Takahashi et al. 1999; Carey and Rutledge 2000; Toracinta et al. 2002), and it is now generally accepted that riming electrification is the main charge separation mechanism of thunderstorms (Williams 1989). Numerical models based on the riming electrification mechanism have successfully simulated the basic characteristics of thunderstorm electrification (Takahashi 1984; Mansell et al. 2005).
Graupel particles are highly charged by collision with ice crystals in cold-chamber laboratory experiments (Reynolds et al. 1957). Both the magnitude and sign of the acquired charge are highly dependent on cloud water content (CWC) and temperature T, as shown in the laboratory experiments of Takahashi (1978). Riming electrification laboratory experiments have yielded many insights into the potential mechanisms by which graupel–ice crystal collisions separate charge under different conditions. However, some incongruities are evident among the reported results (Fig. 1), especially between Takahashi (1978) and several reports using the Manchester cloud chamber (Jayaratne et al. 1983; Saunders et al. 1991; Saunders and Peck 1998).
Takahashi (1978) reported laboratory experiments in which a rotating riming rod (simulating graupel) collided with ice crystals. Temperature and cloud water content greatly affected the charge acquired by the riming rod. Collisions charged the riming rod negatively at temperatures below −10°C across a broad, medium range of cloud water content. Collisions charged the riming rod positively under all other temperature/water content conditions: at temperatures below −10°C with very low or very high cloud water content and at temperatures above −10°C at all cloud water contents.
Substantially different results on charge separation versus temperature and cloud water content were reported in early experiments conducted using the Manchester cloud chamber. Initially, they reported that collisions charged the riming rod positively at high cloud water content and negatively at low cloud water content, with a much smaller temperature dependence than reported previously in Takahashi (1978) (Jayaratne et al. 1983). While they initially reported that in experiments performed at low cloud water content collisions charged the riming rod negatively across the entire temperature range (Jayaratne et al. 1983), they later reported that collisions charged the riming rod positively at temperatures below −20°C at low cloud water content (Saunders et al. 1991).
Gaskell and Illingworth (1980) measured the charge transferred when an artificial hailstone, which was maintained under riming conditions, was impacted with small ice spheres in a wind tunnel in order to simulate in-cloud conditions (Fig. 1). Their cloud water content range was 0.05–0.85 g m−3. At −5° and −10°C, collisions charged the hailstone positively [consistent with Takahashi (1978); consistent with Saunders et al. (1991) only at cloud water content >0.5 g m−3]. At −15°C, collisions charged the hailstone negatively [consistent with Saunders et al. (1991); consistent with Takahashi (1978) only at cloud water content >0.5 g m−3].
Subsequently, three major studies measured charge separation at cloud water content below 4 g m−3 (Pereyra et al. 2000; Berdeklis and List 2001; Takahashi and Miyawaki 2002). The temperature and cloud water content versus charging results of all three show better agreement with Takahashi (1978) (Fig. 1) than with the earlier reports using the Manchester cloud chamber. Pereyra et al. (2000) conducted a series of cloud chamber experiments in Cordoba on riming electrification in which they used a riming target to simulate graupel in a thundercloud. They impacted it with cloud droplets and ice crystals, each of which were grown in their own separate chambers. They reported, “These results are similar to those obtained by Takahashi (1978) and, as has been reported before, are rather different from those obtained in Manchester by Jayaratne et al. (1983), Saunders et al. (1991), and Saunders and Peck (1998)” (Pereyra et al. 2000). Berdeklis and List (2001) measured the charge acquired by a fixed riming rod in a wind tunnel when collided with ice crystals grown in a separate chamber. They reported that the relative humidity in the ice crystal growing chamber has a critical role, with higher relative humidity leading to more negative charging. Takahashi and Miyawaki (2002), to simulate graupel in a thundercloud, floated an ice hemisphere in a wind tunnel and then introduced both ice crystals and cloud drops into the chamber and measured the charge separated by the resulting collisions; the results were consistent with Takahashi (1978). In addition, a later report using the Manchester cloud chamber (Saunders et al. 2006) showed positive electrification of rime at high temperature and low cloud water content; their results were similar to Pereyra et al. (2000).
In a recent report using the Cordoba cloud chamber, Ávila et al. (2013) revisited riming electrification at low cloud water content (Fig. 1). They impacted ice crystals in a wind tunnel on a rime-collecting brass-wire-grid target. At the temperatures tested (from −8° to −12°C), at 3 m s−1 flow speed, the polarity of target charging depended on cloud water content as follows: At −12°C, the riming target charged negatively at 0.05–0.2 g m−3 CWC, and positively at 0.2–0.35 g m−3 CWC. At −10°C, the riming target charged negatively at 0.05–0.3 g m−3 CWC, and positively at 0.3–0.45 g m−3 CWC. At −8°C, the riming target charged positively at 0.2–0.5 g m−3 CWC. In contrast, Takahashi (1978) observed riming target charging to be positive across this entire range of temperatures and cloud water contents. [Here and in Fig. 1, we consider only the results that Ávila et al. (2013) obtained at 3 m s−1 flow speed, in order to make it possible to compare their results with those of Saunders et al. (1991), which were also performed at that flow speed.]
These conflicting laboratory results on riming electrification have hindered the progress of cloud electricity models. To help to discriminate between the laboratory results, riming electrification was examined in field observations of winter clouds in the Hokuriku region of Japan, where riming electrification is the main charging process.
In the Hokuriku region of Japan, heavy winter snow falls from clouds that develop in association with the winter monsoon. As cold air from Siberia (~−13°C) traverses the warm Japan Sea (~10°C), it absorbs sensible heat directly from the sea, and, in addition, as the warmed air rises and cools, it carries with it water vapor from the sea, which releases latent heat as it condenses. Convective cells develop (Fig. 2a). The temperature difference between the ocean and land causes the formation of a low-level convergence, which intensifies cloud convection (Meteorological Research Institute 2005). However, it has been reported that CAPE is less than 150 J kg−1, and updraft is less than 5 m s−1 in Hokuriku winter clouds (Meteorological Research Institute 2005). The general precipitation pattern is for graupel to fall near the coast and for snowflakes and ice crystals to fall inland as these clouds advance. However, during heavy snowfall, convective cells advance farther inland, and graupel falls even farther inland. Although these clouds are as shallow as 4–5 km, they are highly electrified and often produce lightning (Michimoto 1991). It is well known that positive cloud–ground lightning is predominant in Hokuriku winter clouds (Brook et al. 1982). Graupel particles and ice crystals are the major precipitation particles, and lightning is associated with graupel fall (Kitagawa and Michimoto 1994). One advantage to studying Hokuriku winter clouds is that there are no raindrops involved, so both precipitation and cloud electrification processes are simplified.
Space charge distributions and magnitudes in Hokuriku winter thunderclouds (Brook et al. 1982) are similar to those observed in summer thunderclouds (Simpson and Scrase 1937; Williams 1989). In Hokuriku winter clouds, many graupel particles and ice crystals are highly charged, and within each cloud domain they are mostly charged to opposite polarities (Takahashi et al. 1999). Although snowflakes develop intensively in some cloud cells, the space charge carried by them is rather small. Since graupel particles and ice crystals are the main precipitation particles and charge carriers of Hokuriku winter clouds, it is reasonable to hypothesize that electrification results from interaction between these particles (e.g., by riming electrification which results from collision between graupel particles and ice crystals in a riming environment), and this has been confirmed (Takahashi et al. 1999). The space charge carried by graupel and ice crystals in Hokuriku winter clouds was of the same order of magnitude as in summer thunderstorms (Takahashi et al. 1999).
Hokuriku winter clouds were studied using a balloonborne, conjoined, videosonde–hydrometeor videosonde (HYVIS) system. Precipitation particle type, charge, and number concentration were measured by the videosonde component (Takahashi 1990). Cloud water content was derived from images of cloud drops on the scrolling particle collection tape of the HYVIS component (Murakami and Matsuo 1990). Field observations were compared with laboratory experiments.
In Hokuriku winter clouds, cloud water content in most clouds is less than 0.5 g m−3 (Murakami et al. 1994), and cloud-top temperature is usually not very cold (−25°C) because these clouds are not very tall. Therefore, these clouds correspond to the low–cloud water content/high-temperature domain of the riming electrification diagram of Fig. 1. This is also the domain in which the experimental results of Takahashi (1978) and the early Manchester cloud chamber reports differ most.
The sondes launched in this study were composed of two main components: videosonde and HYVIS. Videosondes are designed to measure both the shape and electric charge of precipitation particles in clouds (Fig. 2b; Takahashi 1990). Those used here were of typical design and operate as follows: Particles enter through an inverted cone at the top of the videosonde. The opening is topped with a metal torus to suppress corona discharge. A video camera records particle images, and an induction ring measures their electric charge (detection limit: ≥0.1 pC). An infrared light is mounted above the camera with its beam parallel to the camera’s line of sight. Interruption of this beam by any particle larger than 0.5-mm diameter triggers a flash lamp mounted just above the camera lens. Small particles (0.1–0.5-mm diameter) that are captured in the same frame as the larger, flash-triggering particle are also analyzed. The maximum flash rate is 4 s−1. The induction ring, 70 mm in diameter and 10 mm in height, is mounted at the top of the videosonde. Charge magnitudes are logarithmically amplified, and the detection range is 0.1–200 pC. Charge data are displayed on a set of 10 light-emitting diodes (LEDs) positioned at the bottom of the videosonde within the camera’s field of view (Fig. 2c). The first (leftmost) LED lights up whenever a charged particle passes through the induction ring, and the second LED displays charge polarity (“on” for negative and “off” for positive), with the remaining eight LEDs displaying charge magnitude in binary format. All sensors are installed in an expanded-polystyrene box, which is wrapped in aluminum foil to shield them from external electrical noise. The videosonde transmits data to the ground station via a modulated 1680-MHz carrier wave. Particle charge and space charge values were handled as described in Takahashi et al. (1999) and in Takahashi and Keenan (2004).
Acquiring data by videosonde has certain limitations that must be recognized, and where possible, compensated for (Takahashi et al. 1999; Takahashi and Keenan 2004). The collision of graupel particles and snowflakes with the torus (see below) and the particle breakup that often results are easily detected by the oblique trajectory of the particle or by the appearance of a chain of small ice particles falling obliquely across the field of view. In addition, when a graupel particle bounces off the passage wall (aluminum), an extremely high charge (>200 pC) is recorded along with a complicated charge-wave pattern. These features were used to identify and delete data artifacts of these types.
While it is theoretically possible for corona discharge from the radiosonde to modify precipitation particle charge, this is not expected to be a substantial problem in practice because the radiosonde ascent rate is typically 5 m s−1. Ions ejected from the radiosonde will be left behind when the electric field is less than 50 kV m−1. In thunderstorms, small particles (1–2 mm in diameter) have often been observed to carry charges of a few hundred picocoulombs (Takahashi et al. 1999; Takahashi 2012), and this charge magnitude is much higher than that expected to be generated by ion capture (Whipple and Chalmers 1944). Moreover, as noted previously, the particle entry opening (the inverted cone at the top of the videosonde) is topped with a metal torus to suppress corona discharge.
Particles were classified as raindrop, frozen drop, graupel particle, ice crystal, rimed ice crystal, and snowflake. The videosonde-recorded ice particle images were mostly in good focus, and the several types of particle were categorized by morphology and translucence. Particles categorized as ice crystals were small, thin, and translucent, and a classical crystal shape sometimes could be seen when the crystal face was orthogonal to the camera’s line of sight. The videosonde-imaged ice crystals were of the same shape as ice crystals grown in laboratory chambers in previous experiments [e.g., Fig. 10 in Takahashi (1978)]. Particles categorized as rimed crystals were shaped like ice crystals but opaque. Particles categorized as graupel were lump- or cone-shaped and visibly rimed. Aggregates of two or more ice crystals were categorized as snowflakes. Since graupel particles and ice crystals are the main precipitation particles in Hokuriku winter clouds (see the introduction), they were analyzed as the main components in the charge evolution study.
Cloud droplets (number concentration and droplet size) were observed using the HYVIS system, originally built by Murakami and Matsuo (1990) and later modified by Orikasa and Murakami (1997). It operates as follows: Air is continually drawn at 5 m s−1 through a metal pipe (10-mm diameter), which projects up from the top of the device, topped with an inverted cone. The sampled air flows through the pipe and onto the drop collector: a scrolling, 22 mm wide, transparent film mounted perpendicular to the inlet pipe. Particles >10 μm in diameter strike the film and are collected. The size on film was converted to true drop size (Meteorological Research Institute 2005). A rectangle of dimensions 0.9 mm × 1.2 mm at the center of the film is enlarged 90 times by microscope. After scrolling to a blank strip of film, the scrolling pauses for 10 s, during which time particles are collected and 180 successive frames are transmitted to the ground station. Then, the film scrolls forward to expose a new strip of blank film. The HYVIS device transmits data to the ground station via a 400-MHz carrier wave. Micrographs taken at 2.7, 4.7, and 6.7 s were selected for analysis, with appropriate correction for droplet collection time. To minimize particle overlap, the earliest frame (2.7 s) was used most often. Micrographs were printed and then manually analyzed. Collection efficiencies of particles to film were calculated by Ranz and Wong (1952), and these were applied to our analysis.
In Hokuriku winter clouds, the cloud droplet size distribution is broad (Murakami et al. 1994). The number 6 HYVIS in the present work, a representative example, observed a modal droplet size of 32-μm diameter at 200 m above the cloud base (0.6 km). Cloud water content was calculated from cloud droplets >10 μm in diameter, which is sufficient to accurately measure total cloud water content (Murakami et al. 1994). Below the melting level, melting ice particles were often observed, and when their diameter was less than 100 μm, these droplets were also included in cloud water content measurements. Cloud water content was calculated at ~100-m intervals.
Each conjoined sonde also carried a conventional rawinsonde that measured both temperature and humidity, from which potential temperature and equivalent potential temperature were then calculated.
c. Sonde tracking
Location information from a sonde-mounted GPS device was used to control antenna elevation and azimuth to track the sonde in flight, a procedure derived from the GPS slave method of Suzuki et al. (2012).
d. Other information
A field mill electric field meter was installed at the sonde launch-and-control station. Radar echo images (surface-level PPI) from the Japan Meteorological Agency Niigata radar (80 km northeast of Kashiwazaki) were constantly monitored. The strength of the signal reflected by precipitation particles (radar reflectivity factor) is a measure of its intensity, and it is reported in decibels relative to reflectivity (dBZ). Radar settings were as follows: wavelength of 5.59–5.71 cm; peak power of 250 kW; and 3-dB beamwidth of ~1°. Images were renewed every 5 min. All times are reported in Japan standard time (JST).
3. Observation design
a. Observation protocol
The Hokuriku winter cloud project conducted observations over three periods: (i) from 18 January to 5 February 2010, (ii) from 18 December 2010 to 6 January 2011, and (iii) from 20 to 31 December 2012. Videosonde–HYVIS conjoined sondes were launched only during the third period. Accordingly, we limit our analysis in this paper to data from the third period of observation.
Observations were conducted about 4 km inland from the Sea of Japan at Kashiwazaki (37.19°N, 138.48°E), Niigata Prefecture, in Japan’s Hokuriku region (Fig. 3). A series of 20 videosonde–HYVIS conjoined sondes and two videosondes without HYVIS was launched into Hokuriku winter clouds. Two cold surges struck during the project period, reaching Hokuriku on 24 and 31 December 2012. The first cold surge was the period observed most intensively, with 11 videosonde–HYVIS conjoined sondes launched over 2 days. During each cold surge, the −20°C isotherm dropped in altitude from the 500-hPa level to the 700-hPa level. In addition, during each cold surge, a typical winter monsoon pressure pattern was established: There was intense low pressure over Kamchatka and high pressure over the east of mainland China (Fig. 3). Isobaric lines ran from north to south, and convective clouds lined up along the isobaric lines over the Sea of Japan. These clouds intensified as they reached the Japanese coast (Fig. 3). During the first cold surge, which reached Kashiwazaki on 24 December 2012, average snowfall in the Kashiwazaki area was 14 mm day−1 (24 December 2012; see Fig. 3 for the location of the snow gauges from which the average was calculated), and total (cloud to cloud and cloud to ground) lightning frequency was 49 day−1 (24 December 2012) in the lightning counting domain (36.35°–38.35°N, 137.58°–139.58°E; centered on Kashiwazaki, counted by Franklin Japan Ltd.). Using these same parameters, during the second cold surge, which reached Kashiwazaki on 31 December 2012, average daily snowfall was 33 mm (31 December 2012) and lightning frequency was 69 day−1 (31 December 2012).
Approach of a large radar echo (~10 km wide or larger) initiated operations. Balloons (two 1.2 kg balloons) were prepared for the sonde. Total buoyancy was adjusted to 3–5-kg equivalence depending on snowfall intensity. The videosonde and HYVIS components, firmly mounted to an expanded-polystyrene frame (Fig. 2b), hung about 50 m below the balloons.
With close examination of radar echo movement and electric field information, the sonde was launched into the center of the snow cloud cell, mostly during gusty wind. Precipitation particle shape and charge from the videosonde and cloud drop images from the HYVIS device were monitored on TV screens and recorded to disk. Recording continued until the sonde ascended beyond the cloud top.
Because ice crystals migrate rapidly and widely, it is difficult to infer the site at which they acquire their charge from their observed charge at any given location. Thus, only the graupel charge data were used for analysis of the sites of charge separation.
In this paper, positive graupel charging is analyzed in three ways: 1) positive graupel ratio, 2) charged graupel number frequency, and 3) summed graupel charge. The positive graupel domain was defined using summed graupel charge analysis, the best suited of these analytical methods.
b. The current observations at Kashiwazaki were designed with reference to previous observations of the Hokuriku winter cloud life cycle at Tsunekami and Jōetsu
An earlier videosonde study of winter clouds conducted at Tsunekami and Jōetsu (Takahashi et al. 1999), both also Hokuriku coastal towns, greatly influenced the organization of the present Kashiwazaki observations, and its major findings will be briefly reviewed here. In both Tsunekami and Jōetsu, the electric charge polarity of graupel particles was observed to change greatly with height and cloud life stage (Fig. 4). In the following, each cloud life stage is described by presenting data from a representative sonde flight (see flight number in Fig. 4).
In the Tsunekami/Jōetsu developing stage (T1), in the upper part of the cloud, graupel particles were predominantly negatively charged, and ice crystals were predominantly positively charged. There was an updraft at all cloud levels (5 m s−1 peak value). There were fewer charged particles in the lower part of the cloud, but graupel particles there were predominantly positively charged.
In the Tsunekami/Jōetsu mature stage (T7), videosonde observations found, in addition to negative graupel charging in the upper part of the cloud, evidence for very active positive graupel charging lower in the cloud. Unlike the situation in the developing stage, the updraft (1–2 m s−1) persisted only in the upper part of the cloud, while a downdraft (−3 m s−1) had developed in the lower part of the cloud. The predominant particle species at each level were as follows: in the upper level, mostly positive ice crystals and small, mostly negative graupel particles; in the middle level, highly charged negative ice crystals; and in the lower level, mostly positive graupel particles. These data suggest the following process: The upper level is a major site of graupel formation, and riming electrification there results in negative graupel particles and positive ice crystals. Small negative graupel particles are lofted back up into the updraft area, while large negative graupel particles fall against the updraft. Positive graupel charging takes place in the lower part of the cloud, and the large negative graupel particles that fall from the upper level acquire positive charge as they fall through, eventually becoming positively charged. These newly positive graupel particles join the positive graupel particles that originate in the lower part of the cloud, leading to the large number of positive graupel particles observed there.
These Tsunekami/Jōetsu observations suggest that negative graupel electrification occurs in the upper part of the cloud (low temperature), while positive graupel electrification occurs lower in the cloud (high temperature). The charge polarity reversal temperature of graupel particles was estimated to be −11°C.
In the Tsunekami/Jōetsu dissipating stage (J5), a downdraft (−2 m s−1) was present at all levels, and all particles were descending relative to ground. In the dissipating stage, cloud water content is expected to diminish because of the dominant downdraft. If cloud water content drops too low to support active riming electrification, and if there is no other source of charging, then both negative and positive graupel particles would be expected to fall together with no additional charging during their descent, and the lower-cloud-level positive graupel particle predominance that is seen in the mature stage would be diminished or undetectable in the dissipating stage. This prediction was confirmed by our present observations in Kashiwazaki (sonde numbers 7 and 12).
In Kashiwazaki, typically, clouds completed the developing stage before reaching the sonde launch-and-control station, which was 4 km inland from the coast, and only the mature and dissipating cloud life stages could be observed there.
Nine conjoined-sonde flights (numbers 1, 2, 4, 6, 7, 11, 12, 16, and 21) successfully produced complete data records from both the videosonde and HYVIS components (Table 1). Each was launched into a separate cloud. Sonde number 1 was launched into a layered cloud associated with a frontal system, but since graupel particles were few, this flight was deleted from this analysis. No lightning was detected in this cloud. The rest of the sondes were launched into winter cloud cells that had developed during cold surges. However, sonde numbers 7 and 12 were launched into a cloud cell in the dissipating stage, so those flights were also deleted from this analysis. Thus, six sonde flights remained in the analysis pool (numbers 2, 4, 6, 11, 16, and 21). Although a range of charged graupel particle number concentrations were observed by these six remaining sondes, all observed the same general profile of total graupel charge versus height. All six flights observed positive graupel charging at low cloud levels (high temperature).
During snowfall from cold-surge clouds passing over the sonde station, there were large, sharp swings in electric field to both polarities. The data from all flights are summarized in Table 1. In the following, data from two sondes (numbers 11 and 21), which were launched during the peak of each cold surge, are presented in some detail.
a. Sonde number 11 (1033 JST 24 December 2012)
On 24 December 2012, in the morning, band clouds developed over the ocean near the coast, oriented along the coast (Fig. 5), and echo size at the coast enlarged substantially. Snowfall advanced inland as the main echo cells remained rooted at the coast. Peak radar echo intensity was ~30 dBZ. Equivalent potential temperature changed very little with height within the cloud, evidence of active convection throughout the cloud depth.
A videosonde–HYVIS conjoined sonde (number 11) was launched during graupel fall at 1033 JST 24 December 2012 (Fig. 6a). At launch, there was fog at the sonde launch site, a low hill. Cloud-top height was 3.5 km (−27°C). In the upper part of the cloud (3.0–4.0 km), there was an updraft of 2.7 m s−1. In the layer just below that (2.6–3.0 km), there was a downdraft of −1.2 m s−1. Below 2.6 km, vertical wind velocity was below the detection limit (0.1 m s−1). Ice crystals were abundant throughout the cloud, with peak number concentration of 7 × 103 m−3 at 2.0 km (Fig. 6a; Table 1). In contrast to the continuous distribution of ice crystals, graupel particles were present in two groups: a small group high in the cloud with a local peak at 2.7 km and a large group stretching across the lower half of the cloud with peak graupel particle number concentration of 5 × 102 m−3 at 0.8 km (Fig. 6a; Table 1).
Figure 6b displays particles by altitude and charge. From ~3 to ~1.4 km (from ~−22.5° to ~−11°C), both positive and negative ice crystals were numerous, and positive ice crystals generally outnumbered negative ice crystals throughout that range (Fig. 6b). Below that range, ice crystal numbers decreased. Almost all graupel particles were observed in the lower half of the cloud, and negative graupel particles were more evenly distributed across that range than positive graupel particles. Positive graupel particles were concentrated at ~1.3 km (~−10°C) and below, and they were much more numerous than negative graupel particles (Fig. 6b).
For further analysis, charged graupel particle numbers were plotted in 500-m height bins (Fig. 6c). Above 1.5 km, negative graupel particles outnumbered positive graupel particles. Below 1.5 km, positive graupel particles outnumbered negative graupel particles. From the 1.5–1.0-km bin to the 1.0–0.5-km bin, there was an increase in positive graupel particles and a decrease in negative graupel particles, suggesting the initiation of positive graupel charging at this low cloud level. A similar trend was present in the data from sonde numbers 2 and 4.
Peak CWC in this cloud was 0.4 g m−3 (from CWC values calculated from HYVIS images at ~100-m intervals). The mode of the cloud droplet size distribution within the peak cloud water content bin was 50-μm diameter. The same procedures used here (analysis types, bin size, etc.) were also used to analyze the other videosonde flights, below.
b. Sonde number 21 (0927 JST 31 December 2012)
A typical winter monsoon pattern was present with surface low pressure over Kamchatka and high pressure over eastern China. In the morning, intense cloud bands developed over the ocean perpendicular to the coast. Then they lined up along the coast and approached the launch-and-control station (Fig. 7). Peak echo intensity was 30–35 dBZ. Equivalent potential temperature was nearly constant throughout the cloud depth (Fig. 7).
A videosonde–HYVIS conjoined sonde (number 21) was launched at 0927 JST during graupel fall and strong wind. Three cloud layers were observed within this cloud. Cloud-top height was 4.0 km (−26°C). Vertical wind velocity was +1.0 m s−1 at 4.0–3.4 km, −1.1 m s−1 at 3.4–2.8 km, and +0.6 m s−1 below 2.8 km (Fig. 8a). Later, lightning and thunder were observed.
Ice crystals were widely distributed throughout the cloud (peak number concentration: 2.2 × 103 m−3 at 1.7 km), while most graupel particles were observed in the lower half of the cloud (peak number concentration: 4 × 102 m−3 at 1.5 km) (Fig. 8a; Table 1). Snowflake formation was active throughout the cloud. However, snowflakes were not major charge carriers. As in the cloud observed by sonde number 11, positive graupel particles, compared to negative graupel particles, were much more numerous and had a more bottom-heavy distribution (Fig. 8b).
For further analysis, charged graupel particle numbers were plotted in 500-m height bins. Positive graupel particles clearly outnumbered negative graupel particles in the 2.5–2.0-km bin and below (data not shown). The sonde number 21 charged graupel particle distribution was similar to those from sonde numbers 6 and 16.
Peak CWC in this cloud was 0.5 g m−3 (from CWC values calculated from HYVIS images at ~100-m intervals).
c. Charged graupel particle distribution in the four other clouds (sonde numbers 2, 4, 6, and 16)
The following are descriptions of the four other clouds at the time of observation: The sonde number 2 cloud cell had developed into a large cloud near the coast as a result of vortex formation over Wakasa Bay. The sonde number 4 cloud was a rather isolated cloud cell that had developed close to the balloon launch-and-control station. The sonde number 6 cloud was a large cloud cell near the coast; here, small cells, which had lined up over the ocean, perpendicular to the coastline, then moved in and merged into a larger, preexisting cloud cell over the coast. The sonde number 16 cloud was a large cloud, oriented perpendicular to the coastline, which had developed near the coast.
All charged graupel particles from these four sondes are plotted in Fig. 9. While negative graupel particles were distributed relatively evenly across a wide height range, positive graupel particles were generally concentrated lower in the cloud.
In Hokuriku winter clouds, the riming electrification process is the major charge separation mechanism (Takahashi et al. 1999). According to laboratory experiments, charging depends on not only temperature and cloud water content but also on graupel particle size and on the number and size of ice crystals. In actual clouds, at any given moment, many graupel particles are expected to be relatively uncharged because they are newly formed and/or too small to collide often or effectively with ice crystals, and our data bear this out (e.g., Fig. 6b). Other graupel particles acquire charge. We attempt to determine where in the cloud the graupel particles gain charge and to compare those results with previous laboratory results.
The raw videosonde charged graupel data suggest that positive graupel electrification takes place in the high temperature/low altitude part of the cloud. Electrification was explored further by applying three analytical methods to the data: positive graupel ratio analysis, charged graupel number frequency analysis, and summed graupel charge analysis. These methods analyze the numbers, polarities, charge magnitudes, and locations of charged graupel particles. In addition, summed graupel charge analysis was used to map the zones of negative and positive graupel charging.
a. Positive graupel ratio analysis
Positive graupel ratio analysis plots the proportion of graupel particles that are positively charged versus height (Fig. 10). The ratio of positive graupel particle number to total charged graupel particle number was plotted in 500-m bins. In all cases, the positive graupel ratio was greater than 0.5 at low cloud levels.
b. Charged graupel number frequency analysis
Charged graupel number frequency analysis plots the relative frequency of graupel particles by charge magnitude and polarity at each cloud level (Fig. 11). The graupel particles from all six sondes were combined, and the charge distribution of the graupel particles was plotted for each ascending 5°C bin. The absolute value of the charge of most graupel particles was in the range 1–50 pC. This is the same order of magnitude as was observed in New Mexico summer thunderstorms (Marshall and Winn 1982). From the −15° to −10°C temperature bin, with increasing temperature (decreasing height), the relative frequency of negative graupel particles decreased, and the relative frequency of positive graupel particles increased.
c. Summed graupel charge analysis
The third (and principal) analysis was of summed graupel charge (Fig. 12). The electric charges of all graupel particles, within each 2°C temperature bin, were summed and plotted (with moving average smoothing).
Summed graupel charge analysis can be used to determine where graupel charging occurs. Consider the case of initially uncharged graupel particles falling from high in a cloud. As they collide with ice crystals at high (cold) cloud levels, they incrementally gain negative charge, and the summed graupel charge becomes more negative; this is the negative graupel charging zone. Next, as they continue falling and collide with ice crystals at lower (warmer) cloud levels, they incrementally gain positive charge, and the summed graupel charge becomes more positive; this is the positive graupel charging zone.
In the Kashiwazaki cloud data (Fig. 12), in each cloud, the locations of the positive and negative graupel charging zones were estimated to be the part of each cloud in which there was a trend of positive or negative summed graupel charge, respectively. In each cloud, the zones of negative and positive summed graupel charge and the transition between them were unambiguous, so the zones of positive and negative summed graupel charge were taken as estimates of the positive and negative graupel charging zones.
In each of the six clouds under analysis (see the observations in section 4), there was a transition in summed graupel charge from negative to positive with decreasing height. This was defined as the transition temperature. Hereafter, we refer to the part of the cloud where the summed graupel charge is positive as the positive graupel domain. The transition temperature and the positive graupel domain peak cloud water content were as follows: for number 2, it was −4°C and 0.3 g m−3; for number 4, it was −8°C and 0.4 g m−3; for number 6, it was −5°C and 0.4 g m−3; for number 11, it was −13°C and 0.4 g m−3; for number 16, it was −12°C and 0.5 g m−3; and for number 21, it was −15°C and 0.5 g m−3. The relatively warm transition temperature of cloud number 6 could be a consequence of the unusually high degree of negative charging seen at high altitude in this cloud (see Fig. 9).
It is well known that riming electrification depends greatly on cloud water content (Fig. 1). Consequently, when determining the graupel charge transition temperature in actual clouds, CWC must be high enough to support riming electrification and not so variable as to invalidate the transition temperature determination. The CWC must be adequately high through enough of the cloud, both above and below the transition temperature, to generate enough charged precipitation particles to make accurate analysis possible. Not all clouds will meet these criteria. For example, if a cloud region lacks a sufficiently strong updraft, cloud droplet production will not be fast enough to replace CWC lost to the growth of precipitation particles by accretion and deposition, and there will be a local drop in CWC, complicating further analysis. Of the six clouds under analysis, numbers 4, 11, and 16 had the best CWC profiles (Fig. 12). In contrast, in cloud number 2, CWC remained ≤0.001 g m−3 at altitudes with temperatures from −7° to −11°C. Cloud numbers 6 and 21 each contained a relatively large midcloud region where CWC was ≤0.001 g m−3. When only the three sondes with the best CWC profiles are considered, the mean transition temperature is −11°C, and the mean, peak CWC within the positive graupel domain is 0.4 g m−3. When all six sondes are included, the mean transition temperature is −9.5°C, and the mean, peak CWC within the positive graupel domain is 0.4 g m−3.
If the summed graupel charge in a negative graupel charging zone had been very high, the falling of the negative graupel particles might have been expected to lower the altitude of the apparent boundary between the positive and negative graupel zones. However, in these clouds (Figs. 12a–c), the negative summed graupel charge magnitudes were relatively small and, thus, unlikely to have an effect at the 2°C resolution of our analysis.
The results obtained using positive graupel ratio analysis (Fig. 10) and charged graupel number frequency analysis (Fig. 11), both of which showed a trend toward more positive graupel particles at lower cloud levels, were consistent with the summed graupel charge analysis results. However, summed graupel charge analysis is the best of these methods for mapping the graupel charging zones of these clouds. Positive graupel ratio analysis works best with large populations of charged graupel. If graupel particles are too small to collide with ice crystals or if the graupel supply is limited, as seen in Fig. 6b, the number of charged graupel particles will be too small. Moreover, this method ignores changes in graupel particle charge magnitude that do not change particle polarity. The charged graupel number frequency method also requires large numbers of charged graupel; otherwise, the temperature bins are too broad to allow the transition temperature to be determined with any precision. In contrast, summed graupel charge analysis works well even when the charged graupel particle population is relatively small.
Magono et al. (1983) launched pairs of sondes, a charge sonde followed immediately by a snow crystal sonde, recording snow crystals to investigate the charge carriers of winter clouds in Ishikari, Hokkaido. Their results suggested that positive graupel particles were the predominant charge carriers in the lower part of the cloud and that negative ice crystals were the predominant charge carriers in the upper part of the cloud. The charge sondes observed positive particles below the −12°C level, and they indirectly identified these as graupel particles based on graupel fall trajectory analysis and on data from the subsequent snow crystal sondes. Similarly, they observed negative particles above the −12°C level, and they indirectly identified these as ice crystals.
Takahashi et al. (1999) observed precipitation particle charge and shape more directly using videosondes at Tsunekami and Jōetsu, Hokuriku, as described above (section 3b). Many ice crystals and graupel particles were highly charged and, on average, to opposite polarities. During the developing and mature stages (Fig. 4), when cloud water content is high enough to support active riming electrification, graupel particles in the upper part of the cloud were predominantly negatively charged (peak during the developing stage), and graupel particles in the lower part of the cloud were predominantly positively charged (peak during the mature stage). This is consistent with negative charging of graupel particles in the upper part of the cloud (low temperature) and positive charging of graupel particles in the lower part of the cloud (high temperature). The graupel polarity reversal temperature was estimated to be −11°C. The electric field associated with the space charges carried by graupel particles and ice crystals was high enough to initiate lightning. They proposed that the main charging process was riming electrification. These findings were confirmed in the present work.
Here, at Kashiwazaki, videosonde–HYVIS conjoined sondes recorded not only the shape and charge of precipitation particles but also cloud drop images from which cloud water content was calculated. Thus, the part of the cloud in which positive graupel charging took place could be defined by both temperature and cloud water content. In the three clouds (sonde numbers 4, 11, and 16) with the CWC profiles most conducive to transition temperature analysis (see section 5c), the mean temperature at which summed graupel charge transitioned from negative to positive with decreasing height was −11°C, and the mean, peak cloud water content in the positive graupel domain was 0.4 g m−3. These conclusions are based on summed graupel charge analysis, and they are consistent with the results of positive graupel ratio analysis and charged graupel number frequency analysis.
These Kashiwazaki videosonde–HYVIS results can be used to examine the earlier, contradictory, laboratory results (see the introduction). In the clouds analyzed here, at low cloud water content (≤0.4 g m−3) and warm temperature (≥−11°C), graupel charging was positive. Under these conditions of temperature and cloud water content, Takahashi (1978) reported that graupel (riming rod) charging was positive, while the early Manchester cloud chamber papers reported that graupel (riming target) charging was largely negative (Jayaratne et al. 1983; Saunders et al. 1991; Saunders and Peck 1998). Thus, the present Hokuriku observations show better agreement with the former results. As reviewed in the introduction section and presented in Fig. 1, several subsequent laboratory experiments by several groups have shown greater consistency with Takahashi (1978). A more recent report, using the modified Manchester cloud chamber (Saunders et al. 2006), also shows better agreement with Takahashi (1978) in the warm-temperature/low-CWC region, which they attributed to their having shortened the elapsed time from ice ejection to collision with the riming target. However, many important differences remain, and, as is apparent from this history, it is challenging to recognize and control all of the aspects of laboratory experiments that influence riming electrification. This highlights the importance of frequent confirmation of laboratory results with field observations.
It is not known why laboratory simulation results on riming electrification differ so much between investigators. It may reflect the difficulty of simulating natural conditions. Future laboratory experiments should be done in a large chamber where the wall effect can be avoided, and they should be conducted using free-falling riming targets (simulated graupel) to avoid any effect from the riming target support structure. In the chamber, supersaturation, cloud water content, and cloud droplet size distribution should all be continuously controlled, each factor being critically important to electrification. At the same time, videosonde–HYVIS studies should be extended to other cloud types and regions where temperatures are colder and cloud water contents are larger than in the clouds studied here.
Several individuals made invaluable contributions to this project’s success. Experts from the Central Research Institute of Electric Power Industry (Mitsuharu Nomura and Hiromaru Hirakuchi) helped to set up instruments at the observation site. Students from Yamaguchi University and Kyushu University (Taisei Sasaki, Hidetaka Hirata, Eigo Tochimoto, and Takumi Honda) helped to launch balloons, keep track of sondes, record data, and prepare the early dataset. The authors express their appreciation to Prof. Earle Williams, a reviewer whose excellent comments improved this article. We also thank the two anonymous reviewers for their critical comments. Kanji Takahashi helped to edit the manuscript.