OCTOBER 2007 S T E I G E R E T A L . 3303

Total Signatures of Thunderstorm Intensity over North Texas. Part II: Mesoscale Convective Systems

SCOTT M. STEIGER Department of Earth Sciences, State University of New York at Oswego, Oswego, New York

RICHARD E. ORVILLE AND LAWRENCE D. CAREY Department of Atmospheric Sciences, Texas A&M University, College Station, Texas

(Manuscript received 4 April 2006, in final form 25 January 2007)

ABSTRACT

Total lightning data from the Lightning Detection and Ranging (LDAR II) research network in addition to cloud-to-ground flash data from the National Lightning Detection Network (NLDN) and data from the Dallas–Fort Worth, Texas, Weather Surveillance Radar-1988 Doppler (WSR-88D) station (KFWS) were examined from individual cells within mesoscale convective systems that crossed the Dallas–Fort Worth region on 13 October 2001, 27 May 2002, and 16 June 2002. LDAR II source density contours were comma shaped, in association with severe wind events within mesoscale convective systems (MCSs) on 13 October 2001 and 27 May 2002. This signature is similar to the radar reflectivity bow echo. The source density comma shape was apparent 15 min prior to a severe wind report and lasted more than 20 min during the 13 October storm. Consistent relationships between severe straight-line winds, radar, and lightning storm cell characteristics (e.g., lightning heights) were not found for cells within MCSs as was the case for severe weather in in Part I of this study. Cell interactions within MCSs are believed to weaken these relationships as reflectivity and lightning from nearby storms contaminate the cells of interest. Another hypothesis for these weak relations is that system, not individual cell, processes are responsible for severe straight-line winds at the surface. Analysis of the total lightning structure of the 13 October 2001 MCS showed downward-sloping source density contours behind the main convective line into the stratiform region. This further supports a charge advection mechanism in developing the stratiform charge structure. Bimodal vertical source density distributions were observed within MCS convection close to the center of the LDAR II network, while the lower mode was not detected at increasing range.

1. Introduction placed trailing stratiform (usually displaced to the north in the Northern Hemisphere), and 3) chaotic, unclassi- A mesoscale convective system (MCS) is “a group of fiable mass of convective and stratiform rain. Certain storms that interacts with and modifies the environ- identifiable lightning patterns have been associated ment and subsequent storm evolution in such a way with these MCS types as different dynamics operate in that it produces a long-lived storm system having di- the stratiform and convective regions. Severe weather mensions much larger than individual storms” (large , damaging winds, and tornadoes) occasion- (MacGorman and Rust 1998, p. 258). Houze et al. ally accompanies these systems, mainly in the leading (1990) give a thorough description of the three classic convective line. Lightning has been shown to be a use- MCS reflectivity structures: 1) leading line of deep con- ful indicator of imminent severe weather in supercellu- vection with symmetrically placed trailing stratiform, 2) lar convection (Steiger et al. 2007, hereafter Part I), and leading line of deep convection with asymmetrically it is a major purpose of this study to show lightning’s utility in diagnosing and predicting intense thunder- storm cells within MCSs. Corresponding author address: Dr. Scott M. Steiger, Dept. of Earth Sciences, State University of New York at Oswego, Os- Schuur et al. (1991) and Stolzenburg et al. (1994, wego, NY 13126. 1998) document the electrical charge structure of E-mail: [email protected] MCSs, within both the convective and stratiform re-

DOI: 10.1175/MWR3483.1

© 2007 American Meteorological Society

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC

MWR3483 3304 MONTHLY WEATHER REVIEW VOLUME 135 gions [see Fig. 9 in Stolzenburg et al. (1998) for a con- flashes in the stratiform region are initiated and how ceptual model based on their observations]. They pro- they propagate (e.g., Lang et al. 2004). The relationship pose that two processes are responsible for the charge between reflectivity and lightning will be examined and structure in the stratiform region: growth of mixed- compared to these previous results. phase precipitation particles and subsequent local In MCSs, peak ground flash rates tend to occur later charging in the stratiform region (in situ), and advec- than reports of severe weather (Goodman and tion of charged particles from the convective portion of MacGorman 1986). Cloud-to-ground (CG) flash rates the system into the stratiform region (Rutledge and lag echo volume aloft, which is related to updraft MacGorman 1988). strength in ordinary cells according to Carey and Rut- The convective portion of an MCS analyzed by Carey ledge (1996), and total flash activity in MCSs by 30 min et al. (2005) exhibits a bimodal height distribution of (McCormick 2003). A bipole pattern, the spatial sepa- lightning activity with source peaks at 4.5 km (3°C) and ration between regions of concentrated positive and 9.5 km MSL (Ϫ35°C) while the stratiform region has negative ground flashes, has been noticed in some MCS three peaks at 4.5, 6, and 9 km MSL. These peaks are cases (Orville et al. 1988). Negative ground flashes in thought to be regions of positive charge as the total MCSs are usually associated with the convective region lightning instrument used by Carey et al. [the Lightning while the positive ground flashes concentrate in the Detection and Ranging (LDAR II) research network; stratiform region (Rutledge and MacGorman 1988). same as the one used here] detects more radiation from The average peak ϩCG current in the stratiform region lightning propagating through positive than negative (43 kA) is greater than double the mean peak positive space charge (Rison et al. 1999). The more complex current in the convective region (20 kA) of an MCS charge structure inside the stratiform region agrees studied by McCormick (2003). [Petersen and Rutledge with the results from Stolzenburg et al. (1994, 1998). (1992) and MacGorman and Morgenstern (1998) show Radar reflectivity and lightning source density contours similar results, but there were some exceptions.] The slant downward behind the convective line into the charge regions are more expansive in the stratiform stratiform region (Carey et al. 2005). The excellent region and may provide more charge to a CG flash to agreement between snow trajectories and the slanted produce a larger current (Petersen and Rutledge 1992). path of VHF sources into the stratiform region strongly Sometimes, the convective cells are dominated by posi- supports the charge advection mechanism thought to be tive CG flashes. Typically, the storm that is dominated partly responsible for charging the stratiform regions of by ϩCGs occurs early in the lifetime of the MCS, and MCSs. Carey et al. (2005) caution, however, that in situ is on the southern end of the convective region, where charging cannot be disregarded, as charge advection most severe weather is reported (MacGorman and Rust does not explain the entire charge structure of the 1998, p. 278). stratiform region in the MCS they analyzed. In contrast to the CG flash rates discussed above, As a new cell in an MCS grows and its reflectivity total lightning flashes have shown a better correlation increases, the flash density maximum moves toward the with storm updraft strength (Williams et al. 1999). rear of the cell, eventually merging with lightning in Rapid increases in total lightning rates, or “lightning older cells. McCormick (2003) show a greater number jumps,” can aid in the prediction of severe weather (i.e., of LDAR II flash origins along the back edge of the large hail, severe straight-line winds, and tornadoes) as convective region of a squall line, in the leading edge of they gave 1–15 min of lead time for the storms analyzed the stratiform region, and in the enhanced reflectivity in Williams et al. (1999). “bridge” in the transition zone that connects the con- Mazur and Rust (1983) show that in one storm the vective and stratiform regions. There are relatively few number of cloud flashes was 40 times the ground flash LDAR II flashes associated with the leading edge of count. In contrast to this observation and expectations the convective line and in the stratiform region. LDAR from the MacGorman et al. (1989) elevated charge re- II flashes begin in regions of moderate reflectivity (Ͻ45 gion hypothesis, McCormick (2003) shows intracloud to dBZ) above convective cores at altitudes mostly Ͼ5km cloud-to-ground (IC:CG) flash ratios were at relative MSL, and LDAR II sources are at higher altitudes minimum values (near 12.7) when cells in the 13 Octo- above enhanced reflectivity aloft (indicative of a strong ber 2001 MCS were most intense. This suggests the updraft). LDAR II source density maxima in the con- storm conditions controlling IC and CG flash rates are vective line of the MCS discussed in Carey et al. (2005) more complicated than in the elevated charge hypoth- are well correlated with reflectivity enhancements. esis. Indeed, Mansell et al. (2002) show CG flashes re- These data can also help resolve the issue of where quire the development of a lower charge region (near

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC OCTOBER 2007 S T E I G E R E T A L . 3305 cloud base) of opposite sign to the midlevel charge re- lightning contamination, flashes with positive peak gion to initiate. Another MCS analyzed by McCormick currents less than 10 kA were removed from the NLDN (2003) had a mean IC:CG ratio of 7 and its stratiform dataset (Cummins et al. 1998). The LDAR II source region had greater IC:CG ratios than the convective density (number of sources per unit area per unit region. time interval) was analyzed in horizontal and vertical Unlike many of the aforementioned studies, this projections as described in Part I of this study. The study mostly shows individual storm cell analysis (radar adapted National Aeronautics and Space Administra- reflectivity and total lightning) of three mesoscale con- tion (NASA) flash grouping algorithm (Part I) orga- vective systems (13 October 2001, 27 May 2002, and 16 nized sources into flashes. The total lightning charac- June 2002) using methods similar to those in Part I. The teristics calculated for each storm cell using LDAR II life cycle of MCS storm cells is examined. The radar data within 5, 10, and 20 km of the radar-indicated cell and lightning cell characteristics are related to storm location included the following: the lower quartile, me- intensity and severe straight-line wind occurrence. Cer- dian, 95th percentile (defined as the lightning-based ϩ tain identifiable patterns in the total (IC CG) light- storm top), and modal heights of LDAR II sources; the ning behavior are shown that can be used to infer storm total number of sources within the cylindrical volume; dynamics (e.g., a lightning bow). A comparison be- the number of flashes from sources (total flash rate); tween the lightning and reflectivity structures of the and the IC:CG ratio. convective and stratiform regions of an MCS is shown The radar data to be used in this study were from the to observe how the dynamics, microphysics, and elec- KFWS radar, obtained from NCDC. As in Part I, the trification differ between these two portions of these Warning Decision Support System-Integrated Informa- systems. tion (WDSS-II) software (Hondl 2003) was used to de- termine cell location and characteristics. The radar top (maximum height of the 30-dBZ contour), severe hail 2. Data and methodology index (SHI), and vertically integrated liquid water (VIL) were used to diagnose storm intensity and were Three datasets were used in this study: Dallas–Fort compared with the CG and total lightning characteris- Worth Weather Surveillance Radar-1988 Doppler tics. The radar data were converted to a Cartesian grid (WSR-88D) station (KFWS) data, National Lightning for contouring reflectivity and overlaying lightning Detection Network (NLDN) CG lightning data, and data. LDAR II total lightning data. Analysis was focused on individual convective storm cells within MCSs. Severe Four-dimensional and time–height representations storm reports were obtained from the National Cli- (Figs. 1 and 4) of the radar reflectivity and lightning matic Data Center (NCDC; 2001, 2002a,b). The total data for individual storm cells within MCSs were pro- MCS structure was also examined to compare results duced. Part I of this study details the procedures for with previous studies (e.g., Carey et al. 2005). Partition- creating these plots. For the next section, the storm cell ing convective and stratiform MCS regions was done most closely associated (in time and space) with a se- subjectively by examining the horizontal and vertical vere wind report was chosen. The selected cell was radar structures of an MCS. Convective regions were tracked by using WDSS-II for its lifetime before and defined as having large values of mean reflectivity (Ͼ40 after the report. The cell lifetimes shown in later figures dBZ) and reflectivity gradient, and having reflectivity (Figs. 4–5, 8–9, 11, and 12) were determined from when contours that had a significant bulge upward (large the storm cell identification and tracking (SCIT) algo- slope ϳ0.3; see Fig. 13). Figure 1 from Leary and Houze rithm first identified the cell of interest (first assigned a (1979) shows larger values and greater vertical extent of cell identification number) to when the cell was no reflectivity in the convective region of an MCS. longer identified by that particular identification num- The CG and total (IC and CG) lightning data were ber. This does not necessarily coincide with the cell from the NLDN and the LDAR II network, respec- dissipating completely; this is one of the errors in using tively, and were obtained from Vaisala, Inc., of Tucson, the SCIT algorithm. Also, cell lifetimes were sometimes Arizona. Both of these networks are described in more prematurely ended in the analysis because the cell’s detail in Part I. The following CG lightning character- location exited the analysis domain (30–100-km range istics were analyzed from these data: negative and posi- from the radar; see Part I, section 2). This occurred for tive flash density, percent positive flashes, median peak the 27 May 2002 case. Unlike the majority of the su- current for both polarity flashes, and mean multiplicity percells analyzed in Part I, the MCS cells were not well for each polarity. To decrease the possible intracloud isolated. Hence, it was more likely that radar and light-

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC 3306 MONTHLY WEATHER REVIEW VOLUME 135

FIG. 1. LDAR II source density and mean radar reflectivity of storm cells within an MCS on 13 Oct 2001 (5-min period). (a) Time (UTC)-vs-height display of lightning source density at 5-s and 1-km resolutions. (d) Plan-view projection of mean reflectivity (dBZ, contoured as shown) overlaid on source density (color bar gives values in sources per square kilometer). The resolution of the LDAR II and radar data is 1 km. The asterisk is the location of the cell of interest (at 31, 56 km) detected by WDSS-II. Vertical projections of source density and mean reflectivity for (b) west–east and (e) south–north projections. The height resolutions of the LDAR II and radar data are 1 and 0.5 km, respectively. The color bar is not associated with these panels or the time–height panel. (c) Normalized height histogram of the number of sources and flash origins (shaded) at 1-km intervals. Environmental temperature levels are plotted as bars, and the total numbers of sources, flashes, and peak source height (km) are given above the histogram. The axes are labeled as distances (x: west–east, y: south–north) from the KFWS radar (located at 0, 0 km), and in the vertical direction are heights (km) MSL.

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC

Fig 1 live 4/C OCTOBER 2007 S T E I G E R E T A L . 3307 ning data from other cells infiltrated the analysis radius tilt]. The precipitation was moving outbound at all ver- surrounding the cell of interest. When this contamina- tical levels sampled by the radar at this location. (The tion appeared significant, the cell lightning characteris- cell was located 64 km to the northeast of the KFWS tics were calculated using data within smaller radii (10 radar at this time.) The storm-relative radial velocity in km) of the WDSS-II cell location. the northern portion of the comma head at the lowest tilt was inbound at Ϫ4msϪ1. This radial velocity pat- 3. Lightning and radar reflectivity signatures tern partially explains the comma shape to the reflec- associated with MCS high wind events tivity and the total lightning storm structure observed in Fig. 1d. a. 13 October 2001 The wind report associated with this storm cell oc- Figure 1 shows the mean reflectivity and total light- curred at 0214 UTC. According to Fig. 3, the wind re- ning structure of a storm cell that was part of a leading- port’s time and/or location was inaccurate as it was line, trailing stratiform (LLTS) MCS 14 min prior to a nearly 15 km to the west of the apex in the reflectivity high wind report (26.8 m sϪ1) associated with this cell. and lightning comma-shaped structures [near the apex The storm was at the northern end of an asymmetric is where the strongest surface winds are expected in system (Houze et al. 1990). A total of 41 563 LDAR II these storms (Przybylinski 1995)]. A multimodal source sources composed 540 flashes in this plot, an average of density distribution continued to be observed in this 77 sources per flash. All vertical projections show that storm, with peaks at 4 and 12 km MSL (near 0° and the source density was concentrated in three levels: below Ϫ40°C). Some 42 073 sources composed 392 4 km (above the melting level), 9 km, and 11 km MSL flashes, an average of 107 sources per flash. As with Fig. (T ϽϪ40°C). Distinct multimodal vertical distributions 1, there were three maxima in flash origins at heights of of source density were observed when cells in this study 4, 9, and 12 km MSL. There were two locations in the were within 50 km of the LDAR II network center (this x-height projection where the reflectivity contours ex- cell was 20 km from the network); beyond this distance, tended upward (at x ϭ 42 and 53 km associated with the distributions were unimodal. The LDAR II may have main cell and new cells developing to the southeast of difficulty detecting the lower part of multimodal distri- it, respectively). The two horizontal layers of maximum butions beyond 50 km. [Steiger (2005) shows weaker source density were connected by vertical segments of source powers were associated with the lower peak.] enhanced source density at these locations. There was a The enhanced source density values observed near 10 large region of relatively weak electrical activity ex- km MSL (Ϫ40°C) between x ϭ 6 and 24 km in the tending behind this storm to the west. x-height projection mapped the anvil region of the The time–height history of the reflectivity and total storm. As in Part I, the source density maxima in the lightning structure of this storm is shown in Fig. 4. vertical projections were above near where the reflec- Only data within 10 km of the cell were selected, to tivity contours extended upward (at x, y location 33, 57 reduce contamination from nearby cells. Most of the km; note arrow denoting the reflectivity bulge) indica- cell’s reflectivity contours had their highest extent at tive of where the main storm updraft was located (they 0202 UTC. The 30-dBZ contour was at 14 km MSL at are not as well collocated in the height-y projection, this time. Before this time period, the maximum source Fig. 1e). Figure 2d confirms this, as the source density density was bimodal with values greater than 1600 and maximum in the plan view (located near the cell center) 3200 sources kmϪ1 (5 min)Ϫ1 centered near 5 and 12 km was within the 40-dBZ echo height gradient near the MSL (near Ϫ10° and below Ϫ40°C), respectively. The echo height maximum. upper lightning layer was thicker [cf. the vertical dis- The reflectivity and lightning structure observed with tances between the 1600 sources kmϪ1 (5 min)Ϫ1 con- this storm in the horizontal projection was comma tours in each layer] than the lower one. After 0202 shaped. This radar reflectivity pattern is associated with UTC, there was only one distinct maximum in source strong downdraft winds (Przybylinski 1995). The cell density located at 12 km [1600 sources kmϪ1 (5 min)Ϫ1] center location (indicated with an asterisk in Fig. 1d) and a weak signature of a lower mode at 5 km MSL was within the comma head. Enhanced lightning activ- [800 sources kmϪ1 (5 min)Ϫ1]. The altitudes of both of ity (Ͼ70 sources kmϪ2) was collocated with areas of these lightning layers remained nearly constant large mean reflectivity (Ͼ40 dBZ), unlike with the su- throughout the lifetime of this cell (1 h). The wind re- percells in Part I. Doppler radial wind velocity values port occurred while the altitudes of the upper reflectiv- (not shown) were strong outbound 5–10 km south- ity contours were in descent. The severe wind occurred southeast of the comma head [approximately 30 m sϪ1 while the reflectivity data show the storm was at maxi- (storm-relative radial velocity ϭ 14 m sϪ1) at the lowest mum intensity or while it was weakening. It is difficult

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC 3308 MONTHLY WEATHER REVIEW VOLUME 135

FIG. 2. Same as in Fig. 1 except that the maximum 40-dBZ echo height (km, contoured as shown) is overlaid on the source density in the plan view. The contour interval is 4 km. to diagnose this with better accuracy due to the possible cell’s lifetime. The lower-quartile source altitude, how- timing error in the wind report discussed above. ever, increased from near 5 to 8 km MSL after the wind Lightning heights [95th percentile (lightning top), report. It is interesting to note the upper reflectivity median, lower quartile, and modal source heights] are contours (e.g., 30 dBZ) in Fig. 4 and the lightning top shown to be useful indicators of updraft did not have the same trends after 0202 UTC; the re- strength in Part I. Figure 5 shows little variation in the flectivity contours descended while the lightning top lightning top throughout the 13 October 2001 MCS was steady near 13.5 km MSL. Total flash rates (Fig. 5,

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC

Fig 2 live 4/C OCTOBER 2007 S T E I G E R E T A L . 3309

FIG. 3. Same as in Fig. 1 but at 0214:24–0219:21 UTC 13 Oct 2001. The “W” in each panel represents the time and location of a severe wind report associated with this cell (located at 41, 68 km). bottom) had relative maxima at 0137, 0152, and 0212 mately 7 min before the 13 October 2001 wind report UTC, but showed a general decrease prior to and dur- (Fig. 5). Figure 6 shows ϪCG flash densities were ing the wind report. The last relative maximum flash greater than 0.22 flashes kmϪ2 near the center of the rate of 170 flashes (5 min)Ϫ1 occurred 2 min before the storm (at 37, 58 km) during the time of peak ϪCG report. A relative minimum in the IC:CG ratio oc- activity. This area of large ϪCG flash density coincided curred before the report at 0207 UTC. with an area of large source density values (Ͼ84 sources A peak in negative CG flash rate occurred approxi- kmϪ2). The ϪCG maximum also was located where the

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC

Fig 3 live 4/C 3310 MONTHLY WEATHER REVIEW VOLUME 135

FIG. 4. Time–height plot of the mean radar reflectivity (see color bar at top) and LDAR II source density [sources kmϪ1 (5 min)Ϫ1] for an MCS storm cell between 0129:52 and 0224:18 UTC 13 Oct 2001. To calculate these values, radar and lightning data were averaged within 10 km of the cell centroid during each volume time interval. A severe wind (W) report and the ambient temperature levels are also shown. After 1 source kmϪ1 (5 min)Ϫ1, the source density contour intervals proceed as 2x, where x is the previous contour interval, with the first value being 200 sources kmϪ1 (5 min)Ϫ1. The duration of time between each tick mark is approxi- mately 5 min. vertical projections indicate a bimodal vertical distribu- flash origins was at 9 km MSL (near Ϫ40°C), below the tion of sources (centered at 40, 60 km). source maximum. There were 2910 total sources and 228 flashes included in this plot. This resulted in an b. 27 May 2002 average of 13 sources per flash, significantly less than The convection on 27 May 2002 was an unorganized the average number of sources per flash for the 13 Oc- mass of thunderstorms with stratiform rain and some tober 2001 cells (Figs. 1 and 3). This result was most significant breaks in precipitation. One of these storms likely due to the 27 May 2002 storm’s distance from the produced a severe wind report of 26.4 m sϪ1 between LDAR II network center, being more than double what 2014 and 2026 UTC. Figure 7 shows the reflectivity and it was for the storms shown in Figs. 1 and 3. [Carey et total lightning structure for a volume scan during this al. (2005) discuss how source detection efficiency rap- period. As with the 13 October 2001 wind event, the idly decreases with distance decreasing the number of LDAR II source density had a bow-shaped appearance sources per flash.] The radar data indicated the 27 May in the plan view. A large area of mean reflectivity val- 2002 storm to be more intense than the 13 October 2001 ues greater than 55 dBZ in a south–north orientation cells; maximum contoured reflectivity values were was 5 km to the west of the enhanced lightning activity, greater than 55 dBZ in Fig. 7, while in Figs. 1 and 3 possibly due to the tilt of the cell (see reflectivity dis- these values were between 45 and 50 dBZ. tribution in the west–east vertical projection). There A bow echo in the horizontal reflectivity structure of were two weak echo notches on the backside of this the storm developed approximately 10 min after the storm (which propagated eastward) at positions y ϭ wind report shown in Fig. 7. The total lightning pattern Ϫ24 and Ϫ15 km, but the overall reflectivity pattern continued to have a bow shape, but it was asymmetrical was straight (no bow echo). The high wind report was with maximum source densities within the reflectivity adjacent to the northern weak echo notch. gradient on the northern end (not shown). The signifi- Even though the histogram in Fig. 7 indicates two cant tilting of the main reflectivity core eastward with levels of maximum lightning activity at 10 and 12 km height in the x-height projection in Fig. 7b was a per- MSL (below Ϫ40°C), the central area of the storm cell sistent feature during this storm’s lifetime. Maximum exhibited a unimodal source density distribution in the lightning activity topped the reflectivity core observed vertical projections (as expected since it was 86 km in the west–east vertical projection in Fig. 7b. Note how from the LDAR II network). The maximum number of the source density maximum increased in altitude in the

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC

Fig 4 live 4/C OCTOBER 2007 S T E I G E R E T A L . 3311

during the severe straight-line wind report. The 30-dBZ contour increased from 7 km at 1921 UTC to over 12 km MSL at 2035 UTC; most of this increase was at- tained 30 min before the event. Total lightning activity was unimodal, centered near 10 km MSL (Ϫ40°C) throughout the period. Maximum source densities [Ͼ400 sources kmϪ1 (5 min)Ϫ1] at this altitude occurred during the middle of the report (at 2021 UTC) and were sustained for approximately 15 min after the event. Be- fore 1936 UTC there were no sources detected in this storm below 5 km MSL (near Ϫ5°C), while after this time there was some lightning detected below this level [the minimum contour interval is 1 source kmϪ1 (5 min)Ϫ1]. This occurred just before the development of areas of reflectivity greater than 55 dBZ between 2 and 5 km MSL and descent of 50–55 dBZ to the ground. The storm was also propagating closer to the LDAR II network center during its lifetime (between 118- and 76-km range), and this may have resulted in the detec- tion of more low-level sources. Total lightning and ϪCG flash rates increased to maximum values of 170 and 50 flashes (5 min)Ϫ1, re- spectively, during the 27 May high wind report (Fig. 9, bottom). Negative CG rates significantly increased about 10 min after the development of the low-level FIG. 5. Time history of (top) LDAR II lightning height charac- teristics and (bottom) IC and CG lightning characteristics calcu- maximum reflectivity at 1941 UTC shown in Fig. 8. The lated within 10 km of a 13 Oct 2001 MCS storm cell for each radar total flash rate reached a relative maximum value of volume scan interval. The lightning-based storm top (95th per- greater than 130 flashes (5 min)Ϫ1 at 2001 UTC, coin- centile height: Ltop), modal (ht1), median (med), and lower- cident with the first significant ϪCG flash rate relative Ϯ quartile (quart) heights are shown in the top panel. The CG maximum. Total and ϪCG flash rates both decreased flash rate (pos/neg), percent positive CG flashes (ppos), total flash rate (nfl; right y axis), and IC:CG ratio (rat) are shown in the to relative minimum values at 2006 UTC (approxi- bottom panel. The duration of a severe wind (W) report associ- mately 8 min prior to the wind event) before achieving ated with this storm is also plotted. Each x tick mark represents absolute maximum values during the wind report. This the midpoint of a volume scan and the time interval between each minimum in lightning activity was partly due to a shift tick mark is approximately 5 min. in the areas of large lightning density (LDAR II source and ϪCG flash) farther away from the radar-indicated height-y projection approaching y ϭϪ20 km (the lo- cell location (not shown). The IC:CG ratio for the cell cation of the reflectivity core) from the north in Fig. 7e. had relative maxima at 1951 and 2035 UTC and was As in Part I, the lightning layer was elevated in the nearly an absolute minimum during the wind event. vicinity of the main storm updraft, where the reflectiv- The large ϪCG flash rates during the wind report con- ity contours bulged upward. The thickness of the layer tributed to the low IC:CG ratios (Ͻ5) at this time. Posi- decreased as the distance from the storm center in- tive CG flash rate and percent positive values were creased as well. The regions of relatively thin lightning small throughout this storm’s lifetime. activity in the vertical projections likely show where the Figure 9 (top) shows the evolution of the LDAR II storm anvil was located. source heights for the 27 May 2002 MCS cell. The light- The 27 May 2002 storm intensified before and during ning top was near 11 km MSL (below Ϫ40°C) at 1916 the time it produced severe surface winds. Low-level UTC and then quickly rose to over 13 km during the mean reflectivity values increased from 50 to greater next volume scan. It then decreased to 12 km by 1946 than 55 dBZ at 1941 UTC (Fig. 8). Maximum reflectiv- UTC, before reaching a maximum altitude near 14 km ity values between 50–55 dBZ extended to the ground MSL at 2001 UTC, 13 min before the high wind report. and values greater than 55 dBZ were present near 5 km The lightning top remained elevated during the first MSL during the high wind report. The upper reflectiv- half of the wind report, and then decreased to near 12 ity contours also increased in altitude preceding and km MSL after the report. The other lightning heights,

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC 3312 MONTHLY WEATHER REVIEW VOLUME 135

FIG. 6. Same in Fig. 1 but from 0204:30 to 0209:27 UTC 13 Oct 2001 and instead of the mean radar reflectivity, the negative CG flash density is overlaid on the LDAR II source density in the plan view. The resolution of the CG data is 5 km and the contour interval is 0.08 flashes kmϪ2, starting at 0.06 flashes kmϪ2. Note 1 flash per grid box (25 km2) is 0.04 flashes kmϪ2. The location of the asterisk indicates the cell is located at (39, 60 km). which were lower in elevation and are likely related to did not have changes as drastic as those found with the updraft strength at lower levels in the storm, fol- some supercells (e.g., Part I, Fig. 7). lowed a similar pattern, except they did not ascend at c. 16 June 2002 the initial stage of the storm. They decreased steadily until 1946 UTC and then ascended until peaking at The 16 June 2002 MCS was a well-developed LLTS 2006 UTC (8 min before the wind report). These trends squall line, and the system’s radar and total lightning

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC

Fig 6 live 4/C OCTOBER 2007 S T E I G E R E T A L . 3313

FIG. 7. Same as in Fig. 1 but from 2018:07 to 2023:03 UTC 27 May 2002. The “W” in each panel represents the time and location of a severe wind report associated with this cell. The cell location (asterisk) is (Ϫ42, Ϫ16 km). characteristics are discussed by Carey et al. (2005). The than those associated with the other cells examined in cell analyzed in Fig. 10 was at the southern end of this this study, partially due to this storm’s distance of 108 system. The system was reported to have produced se- km west-southwest from the LDAR II network center. vere straight-line winds of 27.7 m sϪ1 during the time Larger source density values (near 14 sources kmϪ2) shown. The cell was part of a convective region that were located at the northeastern section of the plot; extended in a southwest–northeast orientation. these storms had similar mean reflectivity values as the LDAR II source density values were significantly less cell being examined, but were closer to the LDAR II

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC

Fig 7 live 4/C 3314 MONTHLY WEATHER REVIEW VOLUME 135

FIG. 8. Same as in Fig. 4 but from 1913:47 to 2047:48 UTC 27 May 2002. network centered at (25, 35 km). Enhanced source den- UTC). The total flash rate had two maxima, 55 and 25 sities (4–6 sources kmϪ2) were also associated with the flashes (5 min)Ϫ1, at 0458 and 0538 UTC, respectively. northeastern quadrant (at Ϫ75, 20 km) of the analyzed These peaks had no predictive value of the severe storm. The reflectivity field in the plan view shows a straight-line wind as they occurred after the report. No- significant weak echo notch (WEN in Fig. 10) less than tice how the IC:CG ratios are very low (Ͻ3) throughout 5 km to the northwest of the cell location shown by the the period. The number of ϪCG flashes outnumbered asterisk. The wind report is located 14 km to the north- northeast from this feature in an area of weaker mean reflectivity (30 dBZ), also near another notch that al- most extends through the line, suggesting the wind event may have occurred earlier than the report. As in previous cases, the maximum source density was above where the reflectivity contours extended up- ward at x ϭϪ74 and Ϫ55 km in the x-height projection. However, the reflectivity bulge at y ϭ 5 km had smaller source densities associated with it. There were 13 sources per flash (2475 sources and 186 flashes). Both source density and the number of flash origins peaked at 10 km MSL (near Ϫ40°C). The vertical extent of the mean reflectivity was great- est near the time of the wind report, and showed a general downward trend throughout the period (Fig. 11). The trend in the altitude of the source density con- tours [e.g., 50 sources kmϪ1 (5 min)Ϫ1] was similar. Note the area of 1–50 sources kmϪ1 (5 min)Ϫ1 near 3.5 km MSL at 0503 UTC that coincided with an area of enhanced mean reflectivity (Ͼ50 dBZ). The wind report occurred while the total flash rate was increasing and near a minimum in ϪCG flash rate [15 flashes (5 min)Ϫ1; Fig. 12 bottom]. The ϪCG flash rate reached an absolute maximum at 0448 UTC, ap- proximately 7 min (over one volume scan) prior to the Ϫ wind report. The CG flash rate had a steady decline FIG. 9. Same as in Fig. 5 but for a different storm cell on 27 later in the storm’s lifetime (between 0518 and 0557 May 2002.

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC

Fig 8 live 4/C OCTOBER 2007 S T E I G E R E T A L . 3315

FIG. 10. Same as in Fig. 1 but from 0455:28 to 0500:25 UTC 16 Jun 2002. The time and location of a severe wind report are also shown (W). Note the weak echo notch (WEN). The cell location (asterisk) is (Ϫ79, 11 km). the total flashes between 0443 and 0453 as well as 0513 [Carey et al. (2005, see their appendix C) show relative and 0528 UTC, resulting in IC:CG ratios Ͻ0; this was detection efficiencies Ͻ5% beyond 100-km range from the result of the storm being greater than 100 km from the Dallas-Forth Worth International Airport (DFW) the LDAR II network. The number of sources and total LDAR II network]. The total flash rate trends are also flashes determined from the source data were not rep- suspect as the storm was just coming into the range of resentative of the storm’s lightning production as the network and the initial large increase in total flash source detection efficiency is very small at this range rate may be an artifact of this. Positive CG lightning

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC

Fig 10 live 4/C 3316 MONTHLY WEATHER REVIEW VOLUME 135

FIG. 11. Same as in Fig. 4 but from 0440:37 to 0559:48 UTC 16 Jun 2002. production was insignificant as ϩCG flash rates and in a south–north line along x ϭ 40 km (see plan view percent positive values were less than 3 flashes (5 and x-height projection). Reflectivity contours had a min)Ϫ1 and 8%, respectively. distinct upward protrusion at this position, and large The lightning-based storm top was at a relative mini- source density values through most of the depth of the mum near 13 km MSL (below Ϫ40°C) during the high wind report (Fig. 12, top). The lower-quartile and me- dian heights were also at relative minima near this time. After relative maxima in the heights at 0503 UTC, these values generally decreased until the end of the storm. The lightning top had variations of only 2 km during the cell’s lifetime. The minima in all of the light- ning heights during the wind report and toward the later stages of the storm imply a weaker storm during those time intervals; ϪCG flash rate trends (Fig. 12 bottom) agree with this interpretation. The trend in the cell’s radar top (not shown) was similar to the lightning top. It had a minimum value (7 km MSL) at 0458 UTC near the time of the wind report. The radar top reached a relative maximum altitude at the same time as the lightning top shown in Fig. 12 (0503 UTC).

4. Lightning and radar reflectivity structures in convective and stratiform MCS regions Convective and stratiform areas of an MCS are diag- nosed subjectively in this section; convective regions are defined as having large values of mean reflectivity (Ͼ40 dBZ) and reflectivity gradient and having reflec- tivity contours that extend upward in the vertical pro- jections. Enhanced LDAR II source densities were col- located with maxima in mean reflectivity within the 13 October 2001 mesoscale convective system (Fig. 13).

The source density maxima identified intense storm FIG. 12. Same as in Fig. 5 but for a different storm cell on cells (convection). The most intense cells were oriented 16 Jun 2002.

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC

Fig 11 live 4/C OCTOBER 2007 S T E I G E R E T A L . 3317

FIG. 13. Same as in Fig. 1 but for a total MCS from 0209:27 to 0214:24 UTC 13 Oct 2001. This plot is approxi- mately centered on the LDAR II network located at (25, 35 km), and the horizontal dimensions are 200 km ϫ 200 km (Fig. 1 is 60 km ϫ 60 km). storm were associated with this feature in a vertical the line-normal vertical cross section of an LLTS squall bimodal distribution. The source density contours at- line shown by Carey et al. (2005; see their Fig. 3). An- tained higher altitudes at this position, and then ex- other bulge upward in the reflectivity contours oc- tended downward to the west (rear) of the main line of curred near x ϭ 5 km in the west–east vertical projec- thunderstorms into the stratiform region of the MCS. tion. This was associated with a strong cell located at (5, This corresponds very well with the lightning pattern in Ϫ40 km) in the southern portion of the MCS. The

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC

Fig 13 live 4/C 3318 MONTHLY WEATHER REVIEW VOLUME 135 southern and northern portions of the MCS were more sity maxima were better collocated with large reflectiv- complex than in the central portion; to the north were ity values than was the ϪCG flash density, partially remnants of the supercellular convection that produced because higher spatial resolution (1 versus 5 km for CG a over an hour before the time shown, while to density) was used to plot the source density values. The the south two cells with mean reflectivity values in the reflectivity “bridge” (discussed by McCormick (2003); plan view Ͼ40 dBZ were oriented in a west–east direc- it is a band of enhanced reflectivity values connecting tion. The anvil extended 40 km behind the southern the convective and stratiform rain regions) is observed part of the system in the x-height projection and had extending to the west-northwest from the convective minimal source density values west of the x ϭϪ10 km line at y ϭ 35 km. Enhanced source densities were position. associated with this feature (Fig. 13), but CG activity There were two peaks of source density at 4 and 10 did not occur within the reflectivity bridge during this km MSL in Fig. 13 (near 0° and Ϫ40°C). Three peaks of time period. The reflectivity bridge became wider at flash origins occurred at 4, 8, and 11 km MSL. The two 0304 UTC, and two ϩCG flashes were produced within upper origin peaks were adjacent to the main source this region (not shown). The ϩCG flash density was peak (above and below) and likely corresponded to also closely associated with portions of the convective regions of IC flash initiation while the origin peak at 4 region of the MCS (Fig. 15). Positive CG flash density km MSL was most likely associated with CG lightning values exceeded 0.14 flashes kmϪ2 within the intense production. Proctor (1991) shows the modal CG flash storm cell at (45, 90 km). This cell’s maximum mean origin height near 4 km. More than 113 000 LDAR II reflectivity in the plan view was greater than 50 dBZ, sources and 1352 flashes composed this system, an av- and its large vertical development (the 35-dBZ contour erage of 84 sources per flash while it was near the cen- was above 13 km MSL in both vertical projections) ter of the LDAR II network. indicated this was an intense storm. Surprisingly, there Total lightning activity extended ahead (to the east) were no severe storm reports associated with this in- of the system according to the x-height and plan view tense storm near this time. Unlike the source density in panels in Fig. 13. Weak lightning activity (1–6 sources Fig. 13, ϮCG activity was nonexistent behind (west) kmϪ2) was approximately 20 km east of the main con- and ahead (east) of the convection. The aforemen- vective line between y ϭ 0 and 50 km in areas of mean tioned MCS total and CG lightning patterns persisted reflectivity between 10 and 20 dBZ (plan view). This while the system was within 100-km range of the feature was present in several of the analyzed volume LDAR II network. scans during the lifetime of this system. Weak lightning The lightning top, median, and modal source heights activity also expanded westward into the stratiform re- had little variation throughout the 13 October 2001 gion (mean reflectivity values 10–25 dBZ) after the MCS lifetime (Fig. 16 top). The 95th percentile source convective cells had developed. The plan view in Fig. 13 height was between 13 and 14 km MSL (temperatures is approximately centered on the LDAR II network at below Ϫ40°C). The variations in lightning top were sig- (25, 35 km). The cells farther from the center of the nificantly less than those shown for the 27 May 2002 network to the north and south were of similar strength storm (Fig. 9) and the supercells discussed in Part I. The according to the projections of the mean reflectivity only lightning height that changed significantly during (values and vertical extent), but had significantly dif- the 2.5-h period shown was the lower-quartile height. It ferent total lightning distributions, especially in the ver- was near 9 km at 0122 UTC, then decreased to 6.5 km tical projections. The source density distributions be- MSL at 0152 UTC. The lower-quartile height returned came unimodal as the distance from the network in- to near 9 km MSL by 0302 UTC and remained near this creased, while the closer cells had a distinct bimodal altitude until the end of the period. The lower-quartile appearance. Also, the altitudes of the lower-level height minimum occurred during the same time period source density contours systematically increased with when there was a significant lower region of large distance from the center of the plot, especially in the source density values (e.g., Fig. 13) and a strong bimo- south–north vertical projection. There was less low- dal height distribution signature was present. level lightning activity detected at these greater ranges Total flash rates were between 1200 and 1500 flashes and was the result of the LDAR II system’s inability to (5 min)Ϫ1 until 0307 UTC and then decreased rapidly to detect VHF sources below its horizon. values less than 1000 flashes (5 min)Ϫ1 after this time Negative CG flashes were collocated with areas of (Fig. 16 bottom). Negative CG flash rates were near 300 convection as well (Fig. 14). Negative CG lightning ac- flashes (5 min)Ϫ1 during the earlier stages (first 1.5 h) of tivity was nearly nonexistent in the stratiform region of the system and then decreased significantly concur- the MCS. Figures 13 and 14 show that the source den- rently with the total flash rate. Portions of the MCS

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC OCTOBER 2007 S T E I G E R E T A L . 3319

FIG. 14. Mean reflectivity and ϪCG flash density for a total MCS between 0209:27 and 0214:24 UTC 13 Oct 2001. The color bar (top) gives values of reflectivity in dBZ. The resolutions and contour intervals are the same for the radar and CG flash density as given in Figs. 7 and 6, respectively. Ambient temperature levels are given in the x-height projection. The area of no reflectivity near (0, 0 km) in the plan view represents the volume scan’s cone of silence, while the area of Ͼ55 dBZ reflectivity in the south–north vertical projection below 3 km MSL between y ϭ 70 and 100 km is a measurement error (notice the lack of continuity in the contour intervals on the north end). exited the analysis domain (200 km ϫ 200 km box cen- ally larger than its ϩCG counterpart. Negative current tered on the LDAR II network) after 0312 UTC and values were between 16 and 18 kA while ϩCG currents this was the reason for the significant decrease in flash were between 12 and 14 kA. Approximately 31% rates near this time. The total and ϪCG flash rates had (13%) of the time periods (32 total) analyzed had ϪCG similar trends throughout the period shown; there were (ϩCG) maximum peak currents greater than or equal relative maxima in both at 0127, near 0222, and near to 100 kA. 0307 UTC, and both rates had relative minima near ϩ 0202, 0237, and 0302 UTC. Percent CG values were 5. Discussion and conclusions between 10% and 20%, and the IC:CG ratio had small variations between 2 and 4 during this system’s lifetime. As in Part I, this study examined total lightning be- The system’s ϪCG median peak current was gener- havior in addition to radar reflectivity data to discover

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC

Fig 14 live 4/C 3320 MONTHLY WEATHER REVIEW VOLUME 135

FIG. 15. Same as in Fig. 14 except that the ϩCG flash density is overlaid on the mean reflectivity in the plan view and this system occurred from 0139:46 to 0144:43 UTC 13 Oct 2001. repeatable signatures used to diagnose thunderstorm in- cipitation particles ahead of the parent wind-producing tensity and forecast severe weather (straight-line winds). thunderstorm and lightning initiates and/or propagates The most significant severe straight-line wind signa- into these advected charge regions. The KFWS radar ture found in this study is the development of a light- detected two altitudes near 4 and 12 km MSL of strong ning comma-shape–bow structure in the LDAR II plan outbound radial velocity values [30 m sϪ1 (storm- view source density prior to (over 10 min in the 13 relative radial velocity 14 m sϪ1); not shown] in the 13 October 2001 case) and during severe wind reports October 2001 severe thunderstorm that are near two (Figs. 1–3, 6, and 7). This structure is quite similar to the peak source density heights in Fig. 1c. The low-level radar-detected bow echo, which is associated with velocity maximum is likely associated with the severe strong downbursts (Przybylinski 1995). Two hypotheses surface wind report while the upper one is likely storm- explain the bow feature in source density: 1) new thun- top (environmental winds were greater than 38 derstorm cells develop along the bowing gust front, ini- msϪ1 from the southwest near storm top). The strong tiating lightning in an orientation similar to the front, winds at these levels likely advected the charge regions and 2) the strong winds advect charged cloud and pre- identified by the source peaks in Fig. 1 and played a

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC

Fig 15 live 4/C OCTOBER 2007 S T E I G E R E T A L . 3321

gated along a south–north axis in the plan view. Most of the total lightning activity is ahead (to the east) of the storm reflectivity core. The tilt of the reflectivity core in the west–east vertical projection in Fig. 7b indicates that the cloud and precipitation particles were being advected by the environmental winds. The winds were from the west-southwest at 26 m sϪ1 near 10 km MSL according to the 1200 UTC 27 May 2002 FWD prox- imity sounding (storm-relative environmental winds were west-southwest at 10 m sϪ1). This explains why the maximum LDAR II source densities are to the east of the storm core as the charge region centered at 10 km MSL was also advected on the cloud/precipita- tion particles at this level (which supports the second hypothesis above), but this does not explain the bow shape. Storm advection during the volume scan time interval can also partially explain the storm tilt: by the time the radar beam sampled the upper levels of the storm, it propagated several kilometers downstream (to the east) of where the radar sampled the storm during its first elevation scan. The storm shown in Fig. 7 de- veloped as a series of cells from north to south. A new cell developed to the south (less than 5 km) of the original cell of this storm at 2008:13 UTC (not shown), followed by the development of a new cell to its south by 2023:03 UTC. Each one of these cells developed a

FIG. 16. Same as in Fig. 5 but for the entire MCS that occurred bow structure in the source density and had associated on 13 Oct 2001. These characteristics were calculated using reflectivity bow echoes or weak echo notches within the sources within a 200 km ϫ 200 km box centered on the LDAR II rear portion of the system. Inbound radial velocity val- network for each volume scan. ues near Ϫ20 m sϪ1 (storm-relative radial velocity Ϫ5 msϪ1) indicate rear inflow occurred within these cells. role in the plan-view source density pattern. Examining The evolution of lightning characteristics calculated several radar volume scans before and after the time within 10 km of the radar-identified storm cell must be period shown in Fig. 1 shows the initiation of a new cell examined with caution. Unlike the supercells examined at the apex in the reflectivity structure (the point where in the first section, MCS cells are difficult to isolate. the reflectivity contours are most curved) at 0154:37 Hence, lightning from a nearby cell may contaminate UTC, one volume scan before that shown in Fig. 1. The the volume surrounding the cell of interest. The 10-km mean reflectivity maximum of 45 dBZ at (41, 50 km) in analysis radius was chosen because the 20-km radius Fig. 1 indicates this cell. The first hypothesis discussed ring frequently contained lightning from a nearby cell above is supported by these observations; the strong while the 5-km-radius ring did not include a significant low-level winds initiated a new cell that, in conjunction fraction of lightning associated with the storm. Track- with the original cell, developed a total lightning ing the 13 October 2001 and 27 May 2002 severe wind comma-shape pattern as observed in Figs. 1–3 and 6. producing thunderstorms (Figs. 1 and 7) using WDSS- This lightning pattern persisted for several volume scan II also shows that most of the radar reflectivity of a time intervals (over 20 min). storm is contained within 10 km of the cell location Most of the lightning occurred in the upper levels without significant contamination from other cells. (above 6 km MSL; below Ϫ10°C) of the 27 May 2002 MCS cells develop discretely (along gust fronts) near storm (Fig. 7). The low-level lightning was not detected other cells, and the cell tracking shown by the WDSS-II because the storm was approximately 90 km from the software [using the SCIT algorithm; Johnson et al. LDAR II network (the LDAR II instrument misses (1998)] is likely to have some error in cell identification. radiation from low-level sources mainly due to line-of- A consistent relationship between the radar reflec- sight propagation). In contrast to the 13 October 2001 tivity and lightning characteristics of storm cells within storm, there is only one reflectivity maximum elon- an MCS and the severe straight-line winds they produce

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC 3322 MONTHLY WEATHER REVIEW VOLUME 135 could not be found. Figure 4 shows that the vertical periods as well. A possible explanation for the rela- extent of the 13 October 2001 storm, denoted by the tively constant altitudes of lightning activity is that the maximum heights of the reflectivity contours, began to entire cell lifetime was not captured (SCIT did not de- decrease about 12 min prior to the wind report. How- tect the cell at its early stages and/or prematurely ended ever, the 27 May 2002 storm’s reported wind event oc- its lifetime). For the 27 May case, our methods prema- curred while the storm intensified (vertical extent in- turely ended the cell lifetime as it exited the analysis creased in Fig. 8). Lightning flash rates (total and CG) domain (entered within 30 km from the radar). also had varying relations to severe straight-line wind The inconsistent relationships between storm cell ra- occurrences. Consistent with the decreasing trend in dar and lightning characteristics with severe wind re- the vertical extent of the 13 October 2001 storm, the ports can further be explained by the dynamics respon- total, negative, and positive CG flash rates decreased sible for these strong surface winds. Bow-echo reflec- during the high wind report (Fig. 5). However, the total tivity structures and their associated winds (observed and ϪCG flash rates increased to maximum values dur- on 13 October and 27 May) are thought to be more the ing the high wind report on 27 May 2002 (Fig. 9). The result of system processes, and not necessarily directly increasing trend in total flash rate during the 27 May related to individual convective cell downdrafts. The 2002 storm is partially due to the storm propagating development of a rear-inflow jet [see Wakimoto (2001) closer to the LDAR II network with time from 118- to for an excellent review on the dynamics of MCS high 76-km range (source detection efficiency increased), wind events] that interacts with the convective down- but the large decrease in flash rates after the wind re- drafts of a line of cells produces severe winds at the port in Fig. 9 shows this range effect is not significant surface associated with bow echoes (and lightning enough to mask the true flash rate trends (otherwise source density bows). The cell characteristics calculated the rate would continue increasing as the storm moves in this study (for the detected cells closest in time and closer). If the hypothesis that flash rates are a good space to the wind reports) may not be good indicators measure of storm intensity is valid, the strong relation- of how the storm system, composed of several cells, is ship between the reflectivity development and lightning evolving that is responsible for the severe winds. flash rates between Figs. 8 and 9 also supports that Consistent with the findings of Mazur et al. (1986) these trends are true. The total and ϪCG flash rates and Part I (supercells), where reflectivity contours ex- have opposite trends near the time of the 16 June 2002 tend upward, the layer height and values of maximum high wind report (Fig. 12): total flash rate increases source density are greater (see Figs. 1–3, 7, and 13). while the ϪCG rate decreases. This storm had nearby These features indicate where the main storm updraft is cells contaminating the lightning analysis and there was located. A vertical bridge of enhanced source density a large distance between the cell and wind report (Fig. connects the bimodal distribution of sources in Figs. 10), indicating there was a possible error in its associa- 1–3. The strong updraft at this location enhanced charg- tion with the report. Lightning characteristics are good ing and hence lightning development within the vertical indicators of storm intensity if the MCS cells are iso- column encompassing the updraft. The height-y projec- lated (no other cells within 10 km) as in the 27 May tion in Fig. 7e shows the maximum source density layer 2002 case. The percent positive values were quite small is at a higher altitude within the storm core (at y ϭϪ25 in all cases (Ͻ10%) and the trends have no relation km) relative to the altitude of the lightning layer ex- with severe straight-line wind reports. tending to the north of the core. To test the idea that The lack of variation in the heights of the maximum the altitude of lightning activity can be used to infer LDAR II source densities in Figs. 4 and 8 during each storm strength, linear correlation coefficients between respective time period is significantly different from the the lightning heights [lower-quartile, median, modal, behavior observed during the lifetimes of the tornadic and 95th percentile (lightning top) source altitudes] and supercells in Part I. This seems unreasonable as cell a radar measure of storm intensity (radar top; maxi- updrafts evolve (strengthen and weaken) and the alti- mum height of the storm cell 30-dBZ echo contour) tudes of charge separation and lightning activity should were calculated. The correlation coefficient between respond by ascending and descending (support for this radar top and lightning top for the 27 May 2002 storm hypothesis was shown in Part I). Inspection of the re- is 0.19, but it is not statistically significant at the p ϭ flectivity data along with the SCIT-identified cell cen- 0.05 level. The 13 October 2001 MCS cell has a more troid locations for the 13 October and 27 May storms significant correlation coefficient of 0.41 between these shows that one reflectivity maximum (one cell) was suc- two variables. Surprisingly, the correlation coefficients cessfully tracked for the periods shown. Contamination for the 27 May 2002 storm between radar top and the from other cells was insignificant for most of the time other lightning heights (lower quartile, median, and

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC OCTOBER 2007 S T E I G E R E T A L . 3323 modal heights) are negative (r values near Ϫ0.40). The The weak IC lightning activity to the east of the 13 same coefficients for the 13 October 2001 storm are October 2001 system is associated with the forward an- positive, except for the radar top–lower-quartile source vil of the squall line (Fig. 13). Even though CG flashes height coefficient. These results weakly support the hy- are not associated with this feature, it shows that the pothesis that lightning altitude is a measure of storm cloud is electrified 20 km ahead of the storm. This has cell intensity within an MCS. In contrast, similar analy- an important implication for warning people of danger- sis for supercells in Table 1 of Part I strongly supports ous weather as lightning could propagate to the ground this hypothesis. from the anvil well ahead of the system, and is also Carey et al. (2005) and Dotzek et al. (2005) discuss dangerous to nearby flying aircraft. the reflectivity and total lightning structures of squall Lightning signatures related to storm cell intensity lines that occurred on 16 June 2002 and 7–8 April 2002, within MCSs include larger LDAR II source density respectively, near Dallas–Fort Worth using similar and higher total flash rates as cells strengthen, thicker datasets as in this study. As with the 13 October 2001 and elevated lightning layers associated with inferred system analyzed here, these were LLTS MCSs. An in- updraft regions, and source density comma-shape–bow dividual cell that was associated with a severe wind structures associated with high wind events. Other report from the 16 June 2002 system is analyzed in cases need to be examined to test if these signatures are section 3c. This presents an ideal situation to compare common with these storms and can be used by forecast- results between squall lines from different dates (and ers. The lightning characteristics calculated within 10 seasons) that used the same instrumentation (LDAR II km of an MCS cell show some skill in determining and KFWS WSR-88D). storm intensity and relations with severe straight-line Radar reflectivity and lightning characteristics of the winds. They are not as useful as with supercells (Part I), complete 13 October 2001 squall line have many simi- likely because these cells are not well isolated and in- larities to the results from Carey et al. (2005) and Dot- teract with nearby cells. MCS cells also discretely de- zek et al. (2005). The LDAR II source density is well velop along gust fronts and radar algorithms (i.e., associated with the radar reflectivity in Fig. 13 [also SCIT) have difficulties in tracking them. Finally, storm shown by McCormick (2003) for this case]. The source system and not individual cell processes are thought to density and mean reflectivity contours slope downward be mainly responsible for severe winds associated with to the west of the convective line (the region in the MCSs. These results were only tested with severe west–east vertical projection of Fig. 13 where reflectiv- straight-line winds; improved relationships may occur ity contours have a maximum upward extent identifies with hail and tornado occurrences in MCSs. Analyzing the convective line). Carey et al. (2005) interpret this trends in individual thunderstorm cell characteristics observation to support the charge advection hypothesis show more variation than do total system characteris- for stratiform region charging (Rutledge and MacGor- tics (cf. Figs. 9 and 16). If total lightning data will be man 1988). The in situ charging mechanism, however, used to diagnose/predict intensity changes and severe cannot be rejected (Rutledge et al. 1990; Carey et al.). weather occurrence, it is recommended to use cell The lightning bipole (Orville et al. 1988) is nonexistent analysis. Future research should focus on how to better during the 13 October 2001 case. Figures 14 and 15 isolate lightning with a particular storm cell and study show that both polarity flashes mostly occur within con- other severe weather types in MCSs. vection (this was true for each volume scan during the analyzed system lifetime of Ͼ1.5 h); hence, a separate Acknowledgments. The lightning data (LDAR II and analysis of convective and stratiform CG characteristics NLDN) were obtained from Vaisala, Inc., of Tucson, was not performed. Large values of ϩCG flash density Arizona. We thank Dr. Martin Murphy and Nick De- are associated with a strong convective cell (35-dBZ metriades for their assistance in obtaining the data and reflectivity contour extends above 13 km MSL) in Fig. for many insightful discussions on some of the results. 15. The cell’s position on the northern end of the squall The WSR-88D radar data were initially analyzed using line is unusual as strong storm cells dominated by WDSS-II, and we are grateful to Dr. Valliappa Laksh- ϩCGs typically occur on the southern end (MacGor- manan and Gregory Stumpf for their assistance with man and Rust 1998, p. 278). This cell had strong up- this software. We also wish to thank Brandon Ely for drafts that likely produced large liquid water contents sharing his computer expertise with us in producing that can lead to the growth and positive charging of some of the figures and also for his helpful comments graupel particles (Saunders 1993) in the low levels of on this research. This manuscript was a portion of a the storm and enhanced ϩCG lightning production Ph.D. dissertation and we thank Drs. John Nielsen- (e.g., Williams 2001). Gammon, Donald MacGorman, Fuqing Zhang, and

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC 3324 MONTHLY WEATHER REVIEW VOLUME 135

George Kattawar for their comments. Our research is McCormick, T. L., 2003: Three-dimensional radar and total light- part of a lightning program supported by the National ning characteristics of mesoscale convective systems. M.S. Science Foundation Award ATM-0442011. thesis, Dept. of Marine, Earth, and Atmospheric Sciences, North Carolina State University, 354 pp. NCDC, 2001: Storm Data. Vol. 43, No. 10, 174 pp. [Available from REFERENCES National Climatic Data Center, 151 Patton Ave., Asheville, NC 28801.] Carey, L. D., and S. A. Rutledge, 1996: A multiparameter radar ——, 2002a: Storm Data. Vol. 44, No. 5, 420 pp. [Available from case study of the microphysical and kinematic evolution of a National Climatic Data Center, Federal Building, 151 Patton lightning producing storm. Meteor. Atmos. Phys., 59, 33–64. Ave., Asheville, NC 28801.] ——, M. J. Murphy, T. L. McCormick, and N. W. S. Demetriades, ——, 2002b: Storm Data. Vol. 44, No. 6, 354 pp. [Available from 2005: Lightning location relative to storm structure in a lead- National Climatic Data Center, 151 Patton Ave., Asheville, ing-line, trailing-stratiform mesoscale convective system. J. NC 28801.] Geophys. Res., 110, D03105, doi:10.1029/2003JD004371. Orville, R. E., R. W. Henderson, and L. F. Bosart, 1988: Bipole Cummins, K. L., M. J. Murphy, E. A. Bardo, W. L. Hiscox, R. B. patterns revealed by lightning locations in mesoscale storm Pyle, and A. E. Pifer, 1998: A combined TOA/MDF technol- systems. Geophys. Res. Lett., 15, 129–132. ogy upgrade of the U.S. National Lightning Detection Net- Petersen, W. A., and S. A. Rutledge, 1992: Some characteristics of work. J. Geophys. Res., 103, 9035–9044. cloud-to-ground lightning in tropical northern Australia. J. Dotzek, N., R. M. Rabin, L. D. Carey, D. R. MacGorman, T. L. Geophys. Res., 97, 11 553–11 560. McCormick, N. W. Demetriades, M. J. Murphy, and R. L. Proctor, D. E., 1991: Regions where lightning flashes began. J. Holle, 2005: Lightning activity related to satellite and radar Geophys. Res., 96, 5099–5112. observations of a mesoscale convective system over Texas on Przybylinski, R. W., 1995: The bow echo: Observations, numerical 7–8 April 2002. Atmos. Res., 76, 127–166. simulations, and severe weather detection methods. Wea. Goodman, S. J., and D. R. MacGorman, 1986: Cloud-to-ground Forecasting, 10, 203–218. lightning activity in mesoscale convective complexes. Mon. Rison, W., R. J. Thomas, P. R. Krehbiel, T. Hamlin, and J. Harlin, Wea. Rev., 114, 2320–2328. Hondl, K. D., 2003: Capabilities and components of the Warning 1999: A GPS-based three-dimensional lightning mapping sys- Geophys. Decision Support System-Integrated Information (WDSS- tem: Initial observations in central New Mexico. Res. Lett., II). Preprints, 19th Conf. on Interactive Information Process- 26, 3573–3576. ing Systems (IIPS) for , Oceanography, and Hy- Rutledge, S. A., and D. R. MacGorman, 1988: Cloud-to-ground drology, Long Beach, CA, Amer. Meteor. Soc., CD-ROM, 14.7. lightning activity in the 10–11 June 1985 mesoscale convec- Houze, R. A., Jr., B. F. Smull, and P. Dodge, 1990: Mesoscale tive system observed during the Oklahoma–Kansas PRE- organization of springtime rainstorms in Oklahoma. Mon. STORM project. Mon. Wea. Rev., 116, 1393–1408. Wea. Rev., 118, 613–654. ——, C. Lu, and D. R. MacGorman, 1990: Positive cloud-to- Johnson, J. T., P. L. MacKeen, A. Witt, E. D. Mitchell, G. J. ground lightning in mesoscale convective systems. J. Atmos. Stumpf, M. D. Eilts, and K. W. Thomas, 1998: The Storm Cell Sci., 47, 2085–2100. Identification and Tracking algorithm: An enhanced WSR- Saunders, C. P. R., 1993: A review of thunderstorm electrification 88D algorithm. Wea. Forecasting, 13, 263–276. processes. J. Appl. Meteor., 32, 642–655. Lang, T. J., S. A. Rutledge, and K. C. Wiens, 2004: Origins of Schuur, T. J., B. F. Smull, W. D. Rust, and T. C. Marshall, 1991: positive cloud-to-ground lightning flashes in the stratiform Electrical and kinematic structure of the stratiform precipi- region of a mesoscale convective system. Geophys. Res. Lett., tation region trailing an Oklahoma squall line. J. Atmos. Sci., 31, L10105, doi:10.1029/2004GL019823. 48, 825–841. Leary, C. A., and R. A. Houze Jr., 1979: The structure and evo- Steiger, S. M., 2005: Thunderstorm lightning and radar character- lution of convection in a tropical cloud cluster. J. Atmos. Sci., istics: Insights on electrification and severe weather forecast- 36, 437–457. ing. Ph.D. dissertation, Texas A&M University, 207 pp. MacGorman, D. R., and C. D. Morgenstern, 1998: Some charac- ——, R. E. Orville, and L. D. Carey, 2007: Total lightning signa- teristics of cloud-to-ground lightning in mesoscale convective tures of thunderstorm intensity over north Texas. Part I: Su- systems. J. Geophys. Res., 103, 14 011–14 023. percells. Mon. Wea. Rev., 135, 3281–3302. ——, and W. D. Rust, 1998: The Electrical Nature of Storms. Ox- Stolzenburg, M., T. C. Marshall, W. D. Rust, and B. F. Smull, ford University Press, 422 pp. 1994: Horizontal distribution of electrical and meteorological ——, D. W. Burgess, V. Mazur, W. D. Rust, W. L. Taylor, and conditions across the stratiform region of a mesoscale con- B. C. Johnson, 1989: Lightning rates relative to tornadic vective system. Mon. Wea. Rev., 122, 1777–1797. storm evolution on 22 May 1981. J. Atmos. Sci., 46, 221–250. ——, W. D. Rust, B. F. Smull, and T. C. Marshall, 1998: Electrical Mansell, E. R., D. R. MacGorman, C. Ziegler, and J. M. Straka, structure in thunderstorm convective regions, 1. Mesoscale 2002: Simulated three-dimensional branched lightning in a convective systems. J. Geophys. Res., 103, 14 059–14 078. numerical thunderstorm model. J. Geophys. Res., 107, 4075, Wakimoto, R. M., 2001: Convectively driven high wind events. doi:10.1029/2000JD00244. Severe Convective Storms, Meteor. Monogr., No. 50, Amer. Mazur, V., and W. D. Rust, 1983: Lightning propagation and flash Meteor. Soc., 255–298. density in squall lines as determined with radar. J. Geophys. Williams, E. R., 2001: The electrification of severe storms. Severe Res., 88, 1495–1502. Convective Storms, Meteor. Monogr., No. 50, Amer. Meteor. ——, ——, and J. C. Gerlach, 1986: Evolution of lightning flash Soc., 527–561. density and reflectivity structure in a multicell thunderstorm. ——, and Coauthors, 1999: The behavior of total lightning activity J. Geophys. Res., 91, 8690–8700. in severe Florida thunderstorms. Atmos. Res., 51, 245–265.

Unauthenticated | Downloaded 09/27/21 09:50 AM UTC