X-band Dual-Polarization Radar Observations of the Supercell Storm that Generated an F3 Tornado on 6 May 2012 in Ibaraki Prefecture, Japan

X-band dual-polarization (multi-parameter) radars were used to observe a supercell storm that generated an F3 tornado in Ibaraki Prefecture, Japan on 6 May 2012. The observed data collected for this storm clearly exhibited the typical polarimetric features of a supercell storm, such as the ZDR (differential reflectivity) arc, ZDR column, and the KDP (specific differential phase) column, as well as their time evolution. The ZDR arc emerged at 10 to 15 min before the tornadogenesis. The ZDR column appeared approximately 1 h before the formation of the ZDR arc and was intermittent until tornadogenesis. As the ZDR arc appeared, the column became tall and stable and lasted until the dissipation of the tornado. These ZDR signatures of the supercell storm persisted for approximately half an hour.


Introduction
A tornado is a weather phenomenon that sometimes causes severe damage.On an average, 60 people are killed annually in the United States by tornadoes (Storm Prediction Center, Tornado FAQ, http://www.spc.noaa.gov/faq/tornado/).Significant tornadoes also occur in Japan and result in fatalities and damage to property and infrastructures.Because the majority of significant tornadoes are associated with supercell thunderstorms (Doswell 2001), continuous efforts have been made in the United States to detect supercells to warn about the impending tornadoes.Weather radar observations are essential in detecting the defining characteristics of supercells, such as the reflectivity "hook" echo (Markowski 2002) and the persistent mesocyclone (Stumpf et al. 1998).
Dual-polarization radar, which is increasingly employed for meteorological use, can provide information on the shape and composition of hydrometeors using horizontally and vertically polarized electromagnetic waves.Differential reflectivity, Z DR , is a parameter that is related to the axis ratio of hydrometeors, which can be defined as 10 log (Z H / Z V ) wherein Z H is the horizontal reflectivity factor, and Z V is the vertical reflectivity factor.The parameter Z DR is positive (negative) for oblate (prolate) shapes and is enhanced by large raindrops because they are vertically flattened.The specific differential phase, K DP , is also positive (negative) for oblate (prolate) hydrometeors and is an excellent polarimetric parameter in the X-band for estimating the precipitation rate of heavy rainfalls (Maki et al. 2005).After encouraging initial results, the Japanese Ministry of Land, Infrastructure, Transport and Tourism (MLIT) constructed an X-band dual-polarization radar observation network (X-band polarimetric (multi-parameter) RAdar Information Network, XRAIN) that began operational observations in 2010 (Matsui and Naito 2013).Dual-polarization radar has also been recognized as an efficient tool for discriminating between meteorological and nonmeteorological scatters.Ryzhkov et al. (2005) described the polarimetric analyses of the three tornadic supercell storms and the polarimetric detection of a tornadic touchdown that was associated with lofted debris.Kumjian and Ryzhkov (2008) (hereafter KR08) summarized the polarimetric features of a typical super cell storm.One feature of Z DR is the Z DR arc, which is a shallow region (depth < 2 km) of high Z DR (e.g., > 3 dB) of the supercell that occurs on the right (usually southern) edge of the forward flank downdraft (FFD) near the ground surface.Generally, Z DR increases with Z H (e.g., Marshall and Palmer's raindrop size distribution).However, Z DR can be relatively large in spite of a relatively small Z H in the Z DR arc.Kumjian and Ryzhkov (2009) explained the formation mechanism (of Z DR ) in which size sorting of raindrops is performed in the presence of a vertical increase in speed and veering of the storm-relative wind that favors the development of a supercell.In such an environment, falling small raindrops are advected downstream of the inflow, while large raindrops are retained upstream.Further, the Z DR is enhanced upstream of the low-level inflow.These authors also simulated the Z DR arc using a simple numerical model.Dawson et al. (2015) examined the role of wind shear and showed that the fundamental source of hydrometeor size sorting is the storm-relative wind field and not the wind shear.
Recently, the mechanisms and processes associated with the Z DR arc were examined using detailed numerical simulations.Kumjian and Ryzhkov (2012) reviewed size-sorting mechanisms and quantified the impact of this sorting on the polarimetric radar variables using simplistic bin models.In addition, they examined the inability of simple one-and two-moment bulk microphysics parameterization to reproduce realistic hydrometeor size sorting.Dawson et al. (2014) investigated the impact of size sorting of hail and rain on the qualitative nature of the resulting low-level polarimetric fields in a supercell and revealed that the sorting of the graupel/hail fields had a relatively greater impact on the simulated Z DR arc and the hail core.
Another feature of a supercell is the Z DR column.The column of enhanced Z DR is relatively narrow (4 -8 km wide) in a typical supercell and extends vertically several kilometers above the environmental freezing level that is associated with a strong updraft (KR08).
In this updraft, small raindrops are swept away and large raindrops or water-coated hailstones persist.Sometimes the K DP is also similarly enhanced (K DP column).Snyder et al. (2015) examined the strong spatial association between updrafts and the Z DR columns using numerical simulations with spectral bin microphysics and a polarimetric forward operator and introduced an automated Z DR column detection algorithm designed to provide additional diagnostic and prognostic information pertinent to convective storm nowcasting.
Polarimetric features of supercell storms in the United States have been observed in some detail, with the objective of obtaining a good time resolution.Kumjian et al. (2010) collected rapidly updated (71 -73 s) sectorized, volumetric data from a cyclic supercell storm observed using dual-polarization Weather Surveillance Radar-1988 Doppler (WSR-88D) radar and analyzed the evolution of polarimetric signatures such as the Z DR arc and Z DR column.Tanamachi and Heinselman (2016) also collected observational data of a tornadic supercell using polarimetric WSR-88D radar.These authors described the temporal evolution of polarimetric signatures such as the Z DR column and the tornado debris signature (TDS, Ryzhkov et al. 2005) and discussed possible implications of these rapid data on the warning decision process.
As mentioned above, there have been many studies in the United States pertaining to the observation of polarimetric signatures in supercell thunderstorms.Tornadic supercells have also been reported in Japan (Niino et al. 1993;Suzuki et al. 2000;Mashiko et al. 2009); however, polarimetric data from these supercells have not been collected owing to the lack of operational dual-polarization radar.In contrast to tornadic storms in the United States, F4 or F5 tornadoes have not been reported in Japan.The characteristics of supercell storms in Japan should be investigated to determine whether they have features that are similar to storms observed in the United States.To that end, it is necessary to collect polarimetric data and investigate the signatures in supercell storms in Japan to determine the potential of the polarimetric radar network for the detection of supercell storms and provide the necessary warning information.
On 6 May 2012, an active precipitation band, which was elongated from south to north and moving eastward, produced three tornadoes in Ibaraki and Tochigi prefectures (Japan Meteorological Agency 2012).The tornado at the southern end of the band caused significant damage in Ibaraki Prefecture and was rated as F3 by the JMA.Yamauchi et al. (2013) collected polarimetric data for this storm using a C-band polarimetric radar situated at the Meteorological Research Institute (MRI), JMA that was within 13 -17 km of the tornado.These authors analyzed the tornado vortex and detected the features of a typical supercell including a persistent mesocyclone and a weak echo region (WER), FFD, and TDS.Based on the position of the mesocyclone and the extent of damaged area in their analysis, the authors estimated the duration of the tornado to be from 1235 to 1253 Japan Standard Time (JST = UTC + 9 hours).
The storm was also captured by X-band dualpolarization radars located in the Kanto region.In this study, we report on the observations of the storm and its polarimetric signatures, which were characteristic of supercells, such as those described by KR08.The temporal evolution of features such as the Z DR column and Z DR arc, and their relation with tornado occurrence were investigated to use polarimetric data as a basis for issuing future tornado warnings.

Observations
Figure 1 shows the position and extent of the damage caused by the subject tornado together with the location of the radars used for this analysis and their observation ranges.The Saitama radar is situated at Saitama City and operated by the MLIT as a part of XRAIN and was the radar that was closest to the tornadic storm.This radar is capable of scanning near ground level every 2 min and collecting volumetric data every 5 min.Two plan position indicator (PPI) scans at elevation angles of 1.2° and 2.4° are used for the near-surface scans and 14 PPI scans from 1.2° to 20.0° are used for the volumetric data collection.This radar is not capable of collecting data from the northwest and north owing to the presence of a communication tower that is located beside the radar.The specifications of these radars are listed in Table 1.
The National Research Institute for Earth Science and Disaster Resilience (NIED) also operates two X-band dual polarization radars in Ebina City, Kanagawa Prefecture (EBN radar) and Kisarazu City, Chiba Prefecture (KSR radar).The EBN radar collects data from 12 PPI scans at elevation angles from 0.7° to 10.4° and the KSR radar from 10 PPI scans at elevation angles from 0.7° to 6.9° every 5 min in both cases.A simple attenuation correction was applied to Z H and Z DR using K DP based on the process reported by Suzuki et al. (2010).
The collected data were interpolated onto a Cartesian grid with a horizontal spacing of 0.0045° × 0.0055° (approximately 500 m × 500 m) and vertical grid spacing of 250 m.A correction for advection effects was performed on these data.The location of the observed target was repositioned at the starting time of the volume scan based on the change in the reference frame.The velocity of the moving frame was assumed to be 15 m s −1 to the east and 10 m s −1 to the north, which was estimated by the movement of the echo.
The area that was damaged by the tornado was out of range of the EBN and KSR radars, although they captured the developmental stage of the storm.The tornadic phase of the storm was observed using the Saitama radar.

Results
Figure 2 shows Z H and Z DR observed by the Saitama radar at 1 km above sea level (ASL) at 1235 JST, during the tornado occurrence (Japan Meteorological Agency 2012).Precipitation was strong around the center of the echo wherein Z H exceeded 55 dBZ.The so-called "hook echo" could be clearly observed at the southwestern edge of the echo at approximately 139.93°E and 36.11°N.The Z DR exceeded 4 dB along the south of the Z H 50 dBZ and 30 -40 dBZ contours.Because the echo moved from west-southwest to the east-northeast, the Z DR was high on the right-hand side of the storm, i.e., the inflow edge of the echo.This area of enhanced Z DR was considered to be the Z DR arc.
An east-west vertical section around the center of the hook echo and along the southern edge of the strong precipitation echo at 36.121°N is shown in Fig. 3.The missing areas of data from the storm are the regions where observed data were rejected by a quality control protocol, because of the large value of the Doppler velocity width and the nearly-zero value of (aliased) the Doppler velocity.In Z H (Fig. 3a), the hook echo, which is a hook-shaped horizontal pattern, extended upward from near ground level and continued to the vault with a strong echo at 6 to 7 km ASL above the WER to the east of the hook echo.One distinctive feature of these results was an enhanced Z DR area that also extended upwards at the hook echo and reached more than 3 dB as high as 6 km from ASL (Fig. 3b).This was assumed to be the Z DR column, which was indicative of a strong updraft (KR08).The freezing level at 09 JST was 3.1 km based on the soundings at Tateno, Tsukuba city, Japan.Another area of enhanced Z DR was found to be near the surface just east of the WER, which was the expression of the Z DR arc in the vertical cross-section.
Figure 4 shows the maximum value of Z DR observed by the Saitama radar at 5 km ASL from 1205 to 1300 JST.This period included the time from the tornado genesis (1235 JST) to its dissipation (1253 JST).Because this period was sampled from the volumetric data that were updated every 5 min, this figure corresponds to the temporal change of the Z DR column at 5-min intervals at 5 km ASL.At this altitude, Z DR exceeded 2.5 dB and the Z DR column was detected at 1220 JST.The Z DR increased to 4.0 dB at 1225 JST and the column increased in width at 1230 JST.The column had a ringed shape (KR08) at 1235 JST and passed over the tornado-damaged area from 1235 to 1250 JST.Large values of Z DR were retained until   1255 JST and became unclear at 1300 JST.Development of the Z DR column was also clearly observed at 3 km and 4 km ASL (not shown).ning at 1220 JST.After 1220 JST, the Z DR column was relatively stable until its eventual dissipation.On the contrary, Z DR exceeded 3 dB only at 1150, 1205, and 1215 JST and was less than 3 dB at 1155, 1200, and 1210 JST.This meant that the Z DR column structure of the storm fluctuated or developed periodically before the signatures of a supercell storm became evident.
Figure 7 shows the time variation of Z DR and Z H in east-west vertical sections which is similar to Fig. 5, but the observations were obtained from the EBN and KSR radars from 1120 to 1210 JST.At 1120 JST (Fig. 7a) the Z DR peak associated with the storm was at 139.03°E and was somewhat weak at slightly more than 1 dB.The Z DR gradually increased around 1130 JST (Fig. 7b) and the column began to form at 1140 JST (Fig. 7c).The column matured at 1150 JST with the 3-dB contour reaching nearly 5 km ASL (Fig. 7d).However, the column then decreased in height slightly at 1200 and 1210 JST.Considering the fluctuations of the Z DR at 5 km ASL as shown in Fig. 6, the 5-min time resolution of the observations was apparently insufficient to capture the evolution of the storm; however, the Z DR column was surmised to have developed repeatedly before 1215 JST.
Figure 8 shows the time series of Z H and Z DR at 1 km ASL from 1215 to 1230 JST.At 1230 JST (Fig. 8d), the hook echo and enhanced Z DR area in the southern edge of the echo, the so-called Z DR arc, could be clearly seen as in Fig. 2. The Z DR arc also appeared at 1225 JST to the south of another area of enhanced Z DR (Fig. 8c).Two regions of high Z DR (> 4 dB) became evident, i.e., one inside the 50-dBZ contour of Z H and the other along the 30-dBZ and 40-dBZ contours.The latter was continuous with the Z DR arc shown in Fig. 8(d) and was not clearly seen at 1220 JST (Fig. 8b), except for a small echo around 139.82°E, 36.05°N.At 1215 JST, the area of enhanced Z DR nearly coincided with the high-Z H area, and no Z DR arc was observed.The Z DR arc formed around 1220 or 1225 JST.This development of the Z DR arc was synchronous with that of the Z DR column.
A K DP column was also detected, but this was not as persistent as the Z DR column or the Z DR arc. Figure 9 shows a vertical cross-section of K DP and Z H at 1230 JST.The vertical extent of enhanced K DP (< 4° km −1 ) reached 4 km ASL at 139.87°E.Another region of enhanced K DP below 3 km ASL east of the column around 139.9°E was assumed to be associated with the strong rainfall in a FFD.The column can be clearly seen from 1215 to 1230 JST.The K DP column was located at the northeastern edge of the Z DR column and the offset between the K DP and Z DR columns (approximately 3 km at 1230 JST) was consistent with past studies (KR08; Kumjian et al. 2010).

Summary and discussion
X-band dual-polarization radars were used to collect observational data of a storm that generated an F3 tornado in Ibaraki Prefecture on 6 May 2012.The radar located in Saitama City operated by the MLIT accurately captured the storm data during its tornadic phase, and the radars located in Ebina City and Kisarazu City operated by NIED captured the storm data prior to tornadogenesis.
Typical polarimetric features of a supercell storm, which are often seen in the United States (KR08), such as Z DR arc, and Z DR and K DP columns, were observed throughout the lifetime of the subject tornado.This report is the first account of the polarimetric features of a tornadic supercell storm observed by X-band radars in Japan.The Z DR arc emerged 10 or 15 min before tornadogenesis and lasted for nearly 30 min until dissipation of the tornado.The Z DR column began to develop an hour before tornadogenesis but fluctuated before the Z DR arc appeared.The column also grew tall and became stable 10 or 15 min before genesis of the tornado and lasted nearly 30 min, which was similar to the evolution of the Z DR arc.The K DP column was detected and lasted for approximately 15 min just before tornadogenesis.
The observed signatures of the supercell storm, such as Z DR arc, Z DR and K DP columns, emerged several minutes before the development of the tornado.In particular, the Z DR column was thought to be a potentially good indicator of tornadogenesis because it began to develop before the tornado occurrence and its mature stage closely coincided with the region damaged by the tornado.However, the Z DR column alone is not a direct indication of tornadogenesis.A combination of the development of the Z DR column along with other observable features such as horizontal vorticity associated with a mesocyclone is usually required for a genuine tornado warning.
TDS was also detected in this tornadic storm (Yama uchi et al. 2013).This is a signature of a tornado itself appearing immediately after the tornadogenesis.The information per occurrence of a tornado may be meaningful; however, other information prior to tornadogenesis is essential to ensure the accuracy of a tornado warning.On the contrary, Z DR features accurately represent certain aspects of a supercell and are present before tornadogenesis.
This study is just a single case, and more observations are needed to examine whether these observed features are useful for issuing tornado warnings in Japan.

Fig. 1 .
Fig. 1.Location of the tornado-damage area (solid black rectangle) and positions and observation ranges of the MLIT Saitama radar and the EBN and KSR radars operated by NIED.

Figure 5
Fig. 4. Maximum Z DR (dB, color shading) at 5 km ASL from 1205 to 1300 JST.Labeled times indicate the time when the nearest Z DR peak appeared.The solid black line is the tornadodamage area, as shown in Fig. 1.

2
km east of the column within the Z DR arc.These values increased to 4 dB and the column reached 6 km at 1230 JST.These enhanced Z DR patterns persisted at 1240 and 1250 JST during the tornadic phase and decreased in height by 1300 JST.

Figure 6
Fig. 7. Longitude/height sections of Z DR (dB, color shading) and Z H (contours for 20, 40, and 60 dBZ) from (a) 1210 JST to (f) 1300 JST at 10-min intervals based on EBN and KSR radar data.

Fig. 6 .
Fig. 6.Maximum Z DR (dB, color shading) at 5 km ASL from 1205 to 1300 JST based on EBN and KSR radar data from 1100 to 1225 JST.

Fig. 9 .
Fig. 9. Longitude/height section of K DP (° km −1 , color shading) and Z H (contours for 20, 40, and 60 dBZ) at 1230 JST.Maximum values of K DP and Z H are shown within the latitudinal range between 36.11°N and 36.17°N.Horizontal axis indicates longitude (°E) and vertical axis indicates altitude (m).

Table 1 .
Specifications of the radars