Electrical and Multiparameter Radar Observations of a Severe Hailstorm


Department of Atmospheric Science 

Colorado State University

Ft. Collins, Colorado



Draft of Manuscript Published in the 

Journal of Geophysical Research of the American Geophysical Union


[Carey and Rutledge, 1998, J. Geophys. Res., 103, 13,979-14,000]  

Table of Contents


1. Introduction

2. Measurements and Analysis Methods

3. Observations

    3.1. Meteorological Conditions

    3.2. Storm Evolution: Reflectivity and Storm Reports 

    3.3. Electrical Characteristics Associated with Hail Development in a Supercell

    3.4. Electrical Characteristics Associated with a Severe Hail Storm Complex

    3.5. An Intense Positive Cloud-to-Ground Lightning Cluster

4. Discussion

    4.1 Electrical and Precipitation Characteristics of a Severe Hailstorm

    4.2 Positive Cloud-to-Ground Lightning in Severe Storms: Hypotheses


Appendix A: Multiparameter Radar: Analysis Method

Appendix B: Cloud-to-Ground Lightning: Statistics and Assessment of the Validity of Positive Flashes





    Using data from the 11 cm, CSU-CHILL multiparameter radar, the simultaneous evolution of the radar inferred precipitation structure and electrical characteristics of a severe hailstorm is investigated. We compare the sub-storm point discharge current, intracloud (IC) lightning flash rate, cloud-to-ground (CG) lightning flash rate, ground strike location, and flash polarity to the progression of precipitation types and amounts. This study is unique in that it presents multiparameter radar observations of a thunderstorm complex which exhibited an extremely high IC-to-CG ratio (IC/CG, 20 - 70) and predominantly positive CG lightning (over 74%) after it became severe, producing large hail and weak tornadoes.

    In particular, we investigate the reported relationship between large hail and positive CG lightning. Although a direct correlation is found between a rapid increase in IC/CG, the first positive CG lightning flashes, and the initial production of hail aloft, the temporal and spatial behavior of large hail and positive CG lightning appear to be anti-correlated, as broad peaks in the positive CG flash rate lag relative maxima in the fallout of large hail by up to thirty minutes. The majority of positive ground strikes were adjacent to the main precipitation core in a region of light rain and little or no hail at the surface. Aloft, radar data indicated that ice crystals were aligned vertically in a strong electric field. Corona point observations at the surface indicate that these regions adjacent to the convective core were characterized by net positive charge aloft. Possible mechanisms to explain these observations are discussed.

Table of Contents



1. Introduction

    More than 90% of cloud-to-ground (CG) lightning in warm-season thunderstorms are of negative polarity (Fuquay, 1982; Orville et al., 1987; Reap and MacGorman, 1989; Orville, 1994; Orville and Silver, 1997). As a result, the number of negative CG lightning flashes typically dominates the number of positive CG flashes at any given stage and location in an ordinary (i.e. non-severe), summertime thunderstorm. Exceptions to this situation have been noted in a few scenarios during the warm-season. 1) The dissipating phase (i.e., vertical motions characterized by precipitation filled downdrafts only) of some ordinary thunderstorms can be dominated by positive CG flashes, although the flash density is usually small (e.g., Fuquay, 1982); 2) The trailing stratiform region of mesoscale convective systems (MCS's) (e.g., Rutledge and MacGorman, 1988; Rutledge et al., 1990; Engholm et al. 1990; Stolzenburg, 1990) and the anvil region of some severe storms (Rust et al., 1981a,b) are dominated by positive CG lightning while the deep convective regions of these storms are still characterized by predominantly negative CG lightning. The positive CG flashes in these storms occur over a large area during the course of several hours such that the positive CG flash density is typically low (< 0.01 km-2 h-1, Stolzenburg, 1994); 3) Although cloud-to-ground lightning activity during the mature phase (i.e., characterized by the presence of an active updraft and precipitation filled downdraft; Houze 1993) of non-severe thunderstorms is typically dominated by negative flashes, growing observational evidence suggests that severe thunderstorms which produce large hail (diameter >= 2 cm) and sometimes tornadoes are often characterized by a predominance of positive CG lightning for extended periods of time (>= 30 minutes) during the mature phase (e.g., Rust et. al, 1985; Reap and MacGorman, 1989; Branick and Doswell, 1992; Curran and Rust, 1992; Seimon, 1993; MacGorman and Burgess, 1994; Stolzenburg, 1994). In these severe thunderstorms, the positive CG lightning flashes are clustered in time and space in or near convective regions similar to negative CG flashes in ordinary storms and have comparable flash densities.

    There are currently few detailed case studies of the co-evolving precipitation structure and lightning behavior in severe convection. MacGorman (1993) reviewed earlier studies which suggested that many severe storms are characterized by enhanced intracloud (IC) lightning flash rates and high percentages of positive CG lightning discharges. More recently, Branick and Doswell (1992) and Curran and Rust (1992) both investigated the occurrence of dominant positive CG lightning activity in low-precipitation (LP) supercells which produced large hail and tornadoes. Seimon (1993) analyzed visual and radar observations of a tornadic F5 supercell which produced primarily positive cloud-to-ground lightning during the development phase and which exhibited a reversal in dominant CG polarity from positive to negative at the time of tornado touchdown. MacGorman and Burgess (1994) and Stolzenburg (1994) have surveyed a large number of cases (15 and 24 respectively) of predominantly positive CG lightning producing storms with lightning mapping networks, storm reports, and limited radar data. These studies have confirmed an intriguing and robust one-way correlation between high positive CG flash densities and severe weather. Although not all severe storms produce predominant positive CG lightning, the majority of storms which are characterized by high positive CG flash densities (in excess of negative CG flash densities) during the mature phase are associated with severe weather such as large hail and tornadoes.

    However, the above studies have demonstrated that there is also significant variability in the kinematic, microphysical, and electrical attributes of these storms. As suggested by MacGorman and Burgess (1994), "...the next obvious step in studying positive ground flashes in severe storms is to examine relationships with storm evolution more completely." Therefore, the purpose of this paper is to present detailed observations of the co-evolving precipitation structure, surface electric field, and lightning activity associated with a severe hailstorm occurring in eastern Colorado on 7 June 1995. Over three hours of CSU-CHILL multiparameter radar (S-band, 11 cm) observations with high temporal resolution (<= 6 minutes) and at close range (<= 60 km) are analyzed in the context of the IC lightning flash rate inferred from a flat plate antenna, CG lightning properties such as flash rate, polarity, and ground strike location measured by the NLDN, and the sub-storm point discharge current (proxy for surface electric field) measured beneath the severe hailstorm with a mobile corona point sensor. The location of these instruments relative to the CSU-CHILL radar can be found in Fig. 1. During this three hour period, the storm produced over 74% positive CG lightning (a total of 184 positive flashes) with a peak positive CG flash rate of 7 per minute and a peak percentage of positive lighting of 100% when considering 6 minute averaging periods. In addition, the intracloud lightning flash rate peaked near one per second and the IC/CG exceeded 20. At the same time, the storm produced multiple tornadoes and large hail up to 5 cm in diameter as verified by chase vans working with the CSU-CHILL radar (Hubbert et al., 1997). Multiparameter radar measurables collected by the CSU-CHILL radar allowed inferences regarding hydrometeor types (such as large hail) and precipitation amounts (both rain and hail rates) to be made.


Fig. 1 Depiction of the observational network used in this study. Locations of the 11 cm, dual-polarized CSU-CHILL and Denver WSR-88D radars are indicated. The position of the mobile corona point sensor is shown about 19 km to the southwest of the CSU-CHILL radar. The maximum operating range of the flat plate antenna (located at the CSU-CHILL radar) and the corona point sensor (located about 19 km to the southwest of the CSU-CHILL radar) are depicted by dashed rings of 40 km and 15 km respectively. The location of the closest NLDN (National Lightning Detection Network) LPATS (Lightning Position and Tracking System) station is depicted by a triangle. Hatched areas represent regions of storm echo exceeding 55 dBZ at 1812 and 1952 MDT.


    As a result, we are able to investigate the reported association of positive ground flashes and enhanced intracloud lightning flash rates to the occurrence of severe weather such as large hail using unique multiparameter radar analyses of the three-dimensional precipitation structure. In Sec. 2, we briefly review the data collected and analysis methods used in this study. (More details of multiparameter radar techniques can be found in Appendix A and references given therein). In Sec. 3.1, we discuss the meteorological conditions leading up to the development of a severe storm complex along the Front Range of Colorado. Next, we review the storm evolution using reflectivity data and storm reports in Sec. 3.2. In Sec. 3.3, we present detailed multiparameter radar analyses exploring the temporal and spatial relationship between large hail production, positive CG lightning, the intracloud flash rate, and the surface electric field in an isolated supercell. We then investigate the correlation between large hail, precipitation (rain and hail) rates, and lightning in the entire severe hailstorm complex which resulted from the merging of two supercell storms (Sec. 3.4). In Sec. 3.5, we show the characteristic evolution of the three-dimensional precipitation structure in the vicinity of an intense positive CG lightning cluster. Subsequently in Sec. 4.1, we summarize our observational results and compare and contrast them to past findings. Lastly, we speculate on the plausibility of a few potential hypotheses put forward to explain the dominance of positive CG lightning in Sec. 4.2.

Table of Contents



2. Measurements and Analysis Methods

    Storm precipitation structure was inferred from measurements made by the linearly, dual-polarized CSU-CHILL radar which operates at S-band (2.725 GHz or 11 cm). The radar was recently upgraded with a new antenna and the implementation of a dual-transmit and dual-receive system to improve isolation between the vertical and horizontal channels, allowing accurate estimation of the linear depolarization ratio (Mueller et al., 1995). For the storm under study, a PPI (Plan Position Indicator) sector scanning strategy was employed. Azimuthal spacing provided a slight over-sample with 0.8° while spacing in the elevation angle was variable but software controlled to maintain better than 1 km resolution at 60 km in range. An attempt was made to compromise between satisfactory horizontal and vertical spatial coverage and sufficient temporal resolution. Sector volume times were typically four to six minutes in length and covered the entire extent of the radar echo (at least down to 10 dBZ). However, when the hailstorm approached within 25 km of the radar, it was necessary to sacrifice spatial coverage to maintain acceptable volume times appropriate for deep convection (<= 6 minutes). In this situation, the CSU-CHILL radar maintained horizontal coverage focused on reflectivity cores (> 30 dBZ) and obtained as much vertical coverage as possible in the allotted time. As a result of these occasional scanning deficiencies imposed by the storm location relative to the radar, reflectivity data from the Denver WSR-88D radar was utilized when necessary to supplement analysis of precipitation structure.

    The measured multiparameter radar observables include the horizontal reflectivity (Zh), differential reflectivity (Zdr), linear depolarization ratio (LDR), differential phase (Ydp), and the correlation coefficient at zero-lag (ρdp). One important variable, the specific differential phase (Kdp), was estimated by first isolating the propagation differential phase (fdp) from Ydp by a filtering method (e.g., Hubbert et al., 1993, Hubbert and Bringi, 1995) and then differentiating fdp with range. For complete definitions of the multiparameter radar variables used in this study, we refer the reader to several comprehensive reviews of dual-polarimetric radar measurables (e.g., Rogers, 1984; Bringi and Hendry, 1990; Doviak and Zrnic´, 1993). As discussed in these reviews, multiparameter radar variables provide information on the size, shape, orientation, and thermodynamic phase of hydrometeors in a bulk sense. When used concurrently, these measurables can often be used to categorize the predominant, bulk precipitation type in a radar resolution volume (Jameson, 1985a; Jameson and Johnson, 1990; Doviak and Zrnic´, 1993). For example, it is possible to use the above variables to distinguish large (>= 2 cm) hail from other hydrometeor types (Balakrishnan and Zrnic´, 1990a; Zrnic´ et al., 1993). The technique of bulk hydrometeor identification has been used by various authors to compare the evolution of precipitation structure, especially ice, to that of lightning and storm electric fields (e.g., Carey and Rutledge, 1996; Jameson et al., 1996; Ramachandran et al., 1996; French et al., 1996). In this study, we optimize the above method to distinguish between vertically oriented ice, rain, small hail, large hail, and mixed-phase precipitation (rain mixed with small or large hail). In addition, we have utilized the above variables to calculate both rain and hail rates as discussed in Balakrishnan and Zrnic´, (1990b). For more specific details on how the radar data was processed and analyzed to identify hydrometeor types and estimate precipitation rates, see Appendix A.

    The three-dimensional precipitation structure inferred from multiparameter radar data is placed in the context of storm electricity and lightning data which was obtained from three sources: a flat plate antenna located at the CSU-CHILL radar, a mobile corona point sensor which was located about 19 km southwest of the radar for this event, and the National Lightning Detection Network (NLDN). The placement of these instruments relative to the CSU-CHILL and Denver WSR-88D radars and their range of operation are depicted in Fig. 1.

    The flat plate antenna (e.g., Uman, 1987), or "field change meter," was used to measure the electrostatic field change from both intracloud and cloud-to-ground lightning within about 40 km of the CSU-CHILL radar and was therefore used to obtain the total lightning flash rate. An estimate of the intracloud lightning flash rate was obtained by subtracting the NLDN inferred CG lightning flash rate within 40 km of the CSU-CHILL radar from the flat plate antenna estimated total lightning flash rate. Details regarding the field change meter used in this study and the processing and analysis of the data for lightning flashes can be found in Carey and Rutledge (1996).

    The polarity and strength of the point discharge current was measured with a mobile corona point sensor. The instrument consisted of a conducting rod with a pointed tip which was mounted on a portable instrument tower approximately 5 m AGL and fed into an A-to-D converter sampling at 1 Hz via an operational amplifier. This corona point measured the small currents (microamperes) which flowed from elevated points in response to the net electrostatic field caused by regions of charge within approximately 15 km. As a result, corona current was used to provide a measure of the polarity and relative strength of the near surface electrostatic field due to the net charge within the storm aloft. The gradual and continuous changes (order of a minute) in the corona current were associated with changes in the charge structure aloft due to charge generation and/or advection. Sudden discontinuities (order of a second) in the corona current were associated with changes in the electrostatic field associated with a lightning flash (either IC or CG). More details on corona point sensors and their applications can be found in Engholm et al. (1990) and Carey and Rutledge (1996).

    Cloud-to-ground lightning data including LLP (Lightning Location and Protection) signal strength, polarity, and ground strike location were obtained from the National Lightning Detection Network (NLDN) owned and operated by Geomet Data Services (GDS). The LLP signal strength values were converted to peak current (kA), Ipeak, using the estimate found in Orville (1991),

           Ipeak = 0.19 · ILLP + 2.3       [kA].                                                                       (1)

    The NLDN was recently upgraded (Cummins et al., 1995 and 1996) and now employs a new hybrid lightning location methodology combining time-of-arrival and direction finding techniques (see Holle and López, 1993 for a review of both techniques). Two of the primary objectives of the NLDN upgrade were to 1) improve location accuracy and 2) improve detection efficiency, especially for low peak current events (down to 5 kA). As a result, Cummins et al. (1996) reported an improvement in location accuracy by a factor of 4 to 8, resulting in an average accuracy of 0.5 km, and an estimated flash detection efficiency of 80 to 90%.

    One unexpected result from the recent upgrade of the NLDN is the inclusion of a previously undetected population of small positive discharges. Following changes to sensitivity and wave form discrimination criteria in the NLDN sensors, the percentage of positive flashes has approximately doubled beginning late summer, 1995 (K. Cummins, private communication). These newly detected events are low current (5 to 15 kA estimated peak current) discharges which exhibit flash multiplicities that appear identical to those for larger positive flashes. At the time of this writing, it is not known if these events are long vertical cloud discharges (since they can be detected at distances of up to 500 km and thus falsely detected as positive ground flashes) or if they are a population of previously undetected positive cloud-to-ground discharges. By analyzing and comparing NLDN, flat plate antenna, and corona point sensor data, we demonstrate in Appendix B that the positive cloud-to-ground discharges detected from the storm complex under study are probably not a part of this new subset of potentially false positive CG detections.

Table of Contents



3. Observations


3.1. Meteorological Conditions

    The meteorological conditions of 7 June 1995 along the Front Range of northeastern Colorado (CO) closely resembled those described by Doswell (1980) for High Plains severe thunderstorms. The key ingredients were 1) a quasi-stationary front south of the area through northern New Mexico and the Oklahoma and Texas panhandles, 2) steadily increasing surface dewpoint temperatures in excess of 45° F (7.2° C) with maximum values reaching 53° F (11.7° C) associated with upslope flow which veered from northeasterly to southeasterly during the day at 2 to 5 m s-1, and 3) strong southwesterly flow at 500 mb in excess of 20 m s-1 with weak temperature and vorticity advection associated with a closed low in the Great Basin. At 300 mb, a jet streak rotated through central CO, placing the area of interest in the right exit region which is often a synoptically unfavorable location (i.e., subsident). However, diffluent flow was present at this level which is dynamically favorable for thunderstorm development. Typically, severe weather occurs on several days in succession associated with these relatively benign synoptic conditions (Doswell, 1980). It is interesting to note that severe thunderstorms had occurred on the previous day (6 June 1995) and were therefore expected on 7 June as well.

    Thermodynamic conditions along the Front Range also strongly suggested the possibility of severe weather. A late morning sounding at 1114 MDT (or 1714 UTC; UTC = MDT + 6 h) taken by NCAR (National Center for Atmospheric Research) near Denver, CO (within 20 km of where storms became severe shortly after 1700 MDT) is shown in Fig. 2. The freezing level is located at 2.4 km AGL (4.0 km MSL or 616 mb). The sounding is characterized by an adiabatic environment from the surface to a capping frontal inversion at 680 mb with a relatively dry layer at the surface and a relatively moist layer from about 750 mb to the bottom of the inversion. Above the inversion, the environment was conditionally stable up to 400 mb with the tropopause located at 210 mb and exhibited a deep, mid-level dry layer centered on 500 mb. Due to the presence of this capping inversion, there was substantial Convective Inhibition (CIN) of -725 J kg-1. This cap prevented deep convection prior to 1600 MDT. During the day there was significant moistening in the boundary layer due to the upslope flow. In order to more accurately represent the pre-convective environment which was about five hours following the actual rawinsonde release, the sounding was modified to account for 1) moistening of the surface dewpoint to a maximum of 11.7° C, 2) warming of the surface temperature to a maximum of 22° C, 3) assuming a deep mixed layer up to the Convective Condensation Level (CCL) at about 715 mb, and 4) eliminating the frontal inversion to allow the available convective energy to be explosively released. 


Fig. 2 Skew T-Log P plot of Denver, Colorado radiosonde data from 1114 MDT (1714 UTC) on 7 June 1995. Both temperature and dewpoint temperature are given in degrees Celsius. Wind data, which terminated at 360 mb, is given in m s-1 (1 barb = 10 m s-1).


    Realizing that these assumptions merely allow an estimate of the maximum potential energy in the atmosphere, we obtain a Convective Available Potential Energy (CAPE) of 2268 J kg-1. Winds veered with height from southeasterly to southwesterly and were characterized by moderate shear (20 m s-1) in the lowest 6 km of the troposphere. Combining these shear and buoyancy estimates, the modified sounding yields a bulk Richardson number of 11.3 which is suggestive of supercell development (Weisman and Klemp, 1982; 1984). Considering the favorable synoptic conditions, moderate-to-high CAPE, and moderate shear, the atmosphere was primed for severe thunderstorms capable of producing large hail, damaging winds, and tornadoes.

Table of Contents



3.2. Storm Evolution: Reflectivity and Storm Reports 

    In order to provide an overview of storm evolution in the vicinity of the CSU-CHILL radar during a four hour period, we present a series of two-dimensional (2-D) maximum horizontal reflectivity composites at half-hour temporal resolution in Figs. 3a-i. We chose this approach over a horizontal cut at some arbitrary level because it provides the best overall summary of storm evolution. The 2-D maximum Zh composites were derived from Denver WSR-88D reflectivity data by determining the maximum Zh in a vertical column at every (x, y) location depicted. The Denver WSR-88D radar was used for this purpose instead of the CSU-CHILL because it provides the best overall horizontal and vertical storm coverage with repeating 360° sector volumes from a vantage point well south of the storm complex (cf. Fig. 3). In comparison, the CSU-CHILL was sometimes forced to compromise its scanning in the horizontal and vertical due to the proximity of the storm (see Figs. 3d-g). Surface observations from the public (Storm Data, 1995) and two chase vans working with the CSU-CHILL radar will be set in the context of the storm evolution as seen in these maximum reflectivity composites.


Fig. 3 Two dimensional maximum reflectivity composite computed from Denver WSR-88D radar data. The horizontal (vertical) axis represents east-west (north-south) distance in km from the CSU-CHILL radar at (0,0). The location of the CSU-CHILL radar, Denver WSR-88D radar (7.8, -73.2), and the mobile corona point sensor (-10.6, -15.8) are marked by a 'O'. The large ring centered on the CSU-CHILL radar represents the 40 km maximum range of the flat plate antenna. The 15 km circle centered on the corona point sensor indicates its operational range. Results at a) 1644 (all times MDT). b) 1713. c) 1748. d) 1812 (dashed line at y = -20.5 km represents the plane of the vertical cross-section in Fig. 4. e) 1847. f) 1916. g) 1946. h) 2015. i) 2044. The reflectivity key for all parts (a - i) is located in part a.


    Because of the strong post frontal inversion discussed in Sec. 3.1, convection was suppressed along the Front Range until after 1600 MDT. Between 1600 and 1645 MDT, several non-severe thunderstorm cells developed just north of Denver approximately 60 km to the southwest of the CSU-CHILL, with their echoes topping out at about 9-10 km AGL. As can be seen in the 1700 MDT Denver sounding in Fig. 2, the steering level flow was out of the SSW at about 10 m s-1. As a result, most cells moved generally toward the NNE prior to 1700 MDT. A depiction of these early (1644 MDT) conditions is given in Fig. 3a. Note that peak reflectivities are still generally below 55 dBZ. The largest and most intense cell at this time was centered at about x = -25 km and y = -45 km relative to the CSU-CHILL. From 1644 MDT to 1713 MDT, this particular cell (now centered on x = -15 km and y = -35 km as seen in Fig. 3b) began to develop explosively in the vertical, exhibited early signs of a supercell including a weak echo region and forward overhang, and began to move toward the northeast which is about 14° to the right of the mean shear vector which was directed toward the NNE. This cell which will be called supercell #1 throughout the rest of the paper was just beginning to move within range of the CSU-CHILL flat plate as can be seen in Fig. 3b. Another strong cell of interest which will be referred to as supercell #2 was over northwest Denver by 1713 MDT (x = -37 km, y = -67 km) and began to grow rapidly and exhibit similar signatures of severe supercells.

    By 1748 MDT, both supercell #1 and #2 possessed clearly defined mesocyclones as seen in CSU-CHILL radial velocity data (not shown), bounded weak echo regions (BWER), forward overhangs, and intense (> 65 dBZ) reflectivities as seen in Fig. 3c. At this time, supercell #1 was in a sparsely populated region of Weld County (centered on x = -12 km and y = -23 km) but well within the range of both flat plate and the mobile corona point sensor. As a result, the early development of hail within supercell #1 relative to its electrical characteristics is the subject of the next section. Although no hail was reported by the public for supercell #1 by 1748 MDT, it is clear from reflectivity data that hail was very probable. This will be explored more with multiparameter radar data in the following sections. On the other hand, supercell #2 exhibited a large area of greater than 65 dBZ reflectivities around x = -30 km and y = -47 km over a heavily populated area in the suburbs north of Denver. During this time, there were two public reports of large hail (1.5-1.75 inches or 3.8-4.4 cm). From 1748 MDT to 1812 MDT, supercell #1 moved very little (Fig. 3d) as it produced copious amounts of large hail and a tornado on its right (southern) flank. Public reports of 1 inch hail (2.5 cm) during this period are supported by reports of hail up to 2 inches (5.0 cm) in diameter from two chase vans (one of which contained the mobile corona point sensor as depicted in Fig 1 and Fig. 3d). The other chase van which was outfitted to collect hail and measure rain rates in order to compare to the CSU-CHILL multiparameter radar data (Hubbert et al., 1997) collected several bags of large hail from about 1806 MDT to 1812 MDT. Interestingly, supercell #2 which was located just outside the flat plate's 40 km operational range by 1812 MDT continued to move toward the northeast at about 5 m s-1.

    In order to characterize the vertical structure of supercell #1 near this time (1812 MDT), we present a vertical cross-section of CSU-CHILL reflectivity oriented from east-to-west along y = -20.5 km taken from 1808-1812 MDT (Fig. 4). The dashed line in Fig. 3d delineates the plane of the cross-section. In this vertical cross section, there are several familiar supercell features including a bounded weak echo region (BWER) centered on x = -6 km and 4 < z < 5.5 km, a forward overhang at x = -4 km and 4 < z < 7 km, and an impressive hail cascade (Browning and Foote, 1976) with peak reflectivities near 70 dBZ. To investigate this hail cascade further, we present profiles of the CSU-CHILL multiparameter variables Zdr and ρdp in Fig. 5a and LDR and Kdp in Fig. 5b through the hail cascade marked by the path of "X's" in the vertical cross-section (Fig. 4). Above the melting level (2.4 km AGL), Zh is greater than 60 dBZ, Zdr is slightly negative (-0.2 to -0.4 dB), ρdp is high (0.97 to 0.98), Kdp is less than 0.5° km-1, and LDR is greater than -26 dB. According to the bulk hydrometeor identification method outlined in Appendix A (Table A1), these multiparameter variables indicate the presence of hail. Of course, this is not surprising given the conceptual models of hail growth in supercell thunderstorms (e.g., Browning and Foote, 1976). Below the melting level, the multiparameter variables are consistent with the presence of large hail with Kdp near zero, LDR enhanced to greater than -15 dB, ρdp down to 0.88, strongly negative Zdr (-1 dB), and high reflectivity (> 60 dBZ). Coincidentally, one of the CSU-CHILL chase vans collected hailstones up to 5 cm in diameter within 2 km of the base of the marked hail cascade (Hubbert et al., 1997). The combination of ground truth data (although limited) and the consistency between multiparameter radar inferred areas of hail and conceptual models, provides support for the method of bulk hydrometeor identification outlined in Table A1 (at least for large hail).


Fig. 5  Vertical profile of CSU-CHILL multiparameter radar variables through the hail cascade depicted in Fig. 4.  The melting level is at 2.4 km AGL.  a)  The correlation coefficient (rhv) and differential reflectivity (Zdr, dB).  b)  The specific differential phase (Kdp, ° km-1) and linear depolarization ratio (LDR, dB).


Fig. 4  East-west vertical cross-section of CSU-CHILL horizontal reflectivity (dBZ) at y = -20.5 km (see Fig. 3d). Note the forward overhang and bounded weak-echo region (BWER) from x =-7 km to x = -3 km. The path of the hail cascade which is analyzed in Figs. 5a,b is marked by X's.


    Over the next half hour, observations from the corona point van included large hail and later (after 1820 MDT) a prodigious amount of small hail mixed with heavy rain which resulted in flooded farm fields. The large hail and heavy rain produced over $33 million dollars of damage in several small towns and surrounding fields in Weld County to the southwest of the CSU-CHILL radar (Storm Data, 1995). There were four public reports of large hail (2-5 cm) and one tornado report between 1812 MDT and 1847 MDT for the two supercells depicted in Fig. 3e which were beginning to merge into one large thunderstorm complex as supercell #2 caught up with supercell #1 from the south. By 1916 MDT, the two supercell storms were completely merged at mid- to upper-levels and could be distinguished only by two reflectivity maximum (> 65 dBZ) near the surface that are evident in the maximum reflectivity composite in Fig. 3f (supercell #1: x = 8 km, y = -4 km; supercell #2: x = 12 km, y = -20 km). This merged hailstorm complex was moving almost due east by this time. During the period from 1847 MDT to 1916 MDT, there were three more public reports of large hail (2.5-4.4 cm) and a tornado near Kersey, CO (x = 2.2 km and y = -9.6 km). This tornado which was evident in the CSU-CHILL radial velocity data as a tornado vortex signature (TVS; Lucci, 1996) appeared to occur along a shear zone in a descending hail curtain adjacent to the storm updraft. Lucci (1996) analyzed and compared the kinematic structure of the hailstorm complex during this period to the occurrence of positive CG lightning. Note that a third cell began to form by 1916 MDT on the right flank (southwest side) of the merged supercell #1 and #2.

    From 1916 MDT to 1946 MDT, the merged complex continued to move toward the ENE, producing large hail and a weak tornado in Weld county to the ESE of the radar. Notice that the reflectivity maximum (> 65 dBZ) originally associated with supercell #1 at 1916 MDT (Fig. 3f: x = 8 km, y = -4 km) has fallen out of the storm by 1946 MDT as seen in Fig. 3g. As will be demonstrated in Sec. 3.5 below, the fallout of large hail in this area around 1929-1940 MDT was followed by an intense positive CG lightning cluster (grouping of 20 ground strike locations in a 10 km x 10 km area, centered on x = 22 km and y = 15 km, during an 18 minute period). Interestingly, this positive CG cluster which occurred from about 1940 to 2001 MDT was followed closely by a public report of two tornadoes at 2004 MDT only 10 km to the SSE of the positive CG cluster.

    As the storm complex continued to move toward the ENE from 2015-2044 MDT (Figs. 3h,i), it took on the characteristics of a short, severe multicell squall line with new cells developing on the right (southwestern) flank. Occasional reports of large hail continued during this period. After 2044 MDT, the storm continued its eastward march and eventually moved out of range for quality CSU-CHILL multiparameter radar analyses (best at ranges of <= 60 km). With the storm evolution described in general for this four hour period, it is now possible to analyze in detail the relationships between radar derived precipitation structure and electrical behavior for this anomalous hailstorm.

Table of Contents



3.3. Electrical Characteristics Associated with Hail Development in a Supercell

    Before tackling the association between hail development and electrical characteristics for the complex storm evolution of the two merging supercells discussed in the previous section, we first investigate these relationships for supercell #1 alone from its genesis at about 1620 MDT up to its merger (of the 55 dBZ reflectivity contours) with supercell #2 around 1855 MDT (e.g., Figs. 3e,f). The reasons behind this analysis choice are two fold: 1) any correlation between electrical characteristics and precipitation development will be more evident when analyzing a single supercell, and 2) simultaneous corona point sensor data and flat plate antenna data were available for supercell #1 as it approached the CSU-CHILL from the southeast while supercell #2 was still out of range (e.g., Figs. 3c-d) and therefore not affecting the corona point or flat plate measurements.

    The evolution of intracloud and cloud-to-ground lightning for supercell #1 is shown in Fig. 6. An estimate of the IC flash rate is only available while supercell #1 was fully within range of the flat plate antenna and while supercell #2 was still out of range (1730-1810 MDT, e.g. Figs. 3b-d). Note that this particular cell produced exclusively negative CG lightning in its early, non-severe phase. The reflectivity structure in Fig. 3a. at 1644 MDT just follows this maxima in the negative CG flash rate from 1630-1645 MDT. Interestingly, as this cell began to develop rapidly in the vertical, it produced no NLDN detectable CG lightning. This dearth in CG lightning lasted for just over one hour (1645-1750 MDT) as the cell evolved from an ordinary thunderstorm to a severe supercell. Based on flat plate measurements, we know that the storm was indeed electrified because it was producing IC flashes at a rate of 2.6 - 4.4 flashes min-1 from 1730-1750 MDT (Fig. 6). When the cell finally produced CG lightning again at 1750 MDT, the polarity switched to primarily positive and the number of IC flashes began to increase dramatically as can be seen in Fig. 6. Supercell #1 continued to produce primarily positive CG lightning (87%) for the rest of its lifecycle (1750-1900 MDT) until it merged with supercell #2. (As shown in the next section, the merged storm also produced mostly positive CG lightning flashes). A broad peak in the positive CG flash rate for supercell #1 began at 1835 MDT and continued until the merger process commenced at 1900 MDT. This peak positive flash rate was actually fairly low (1.2 flashes min-1) compared to typical negative CG flash rates in ordinary storms.


Fig. 6 Evolution of the number of intracloud (IC), positive cloud-to-ground (+ CG), and negative cloud-to-ground (- CG) lightning per five minute period for supercell #1 from 1620 - 1900 MDT. An estimate of the IC lightning flash rate was only available from 1730 - 1815 MDT (see text for explanation).


    In Fig. 7, the precipitation history at the surface for the same time period covered by Fig. 6 is given. In particular, the evolution of the rain and hail mass fluxes and the peak hail rate calculated from Kdp and Zh (as outlined in Appendix A) is depicted for supercell #1 from 1620 MDT to 1900 MDT. First, note that a peak in the rain mass flux and peak hail rate occurred during and just following the peak in the negative CG flash rate from 1635 MDT to 1645 MDT. Most ordinary thunderstorms exhibit a similar correlation between the rain mass flux, hail rate, and negative CG flash rate (e.g., Carey and Rutledge, 1996) as seen in this non-severe precursor phase of supercell #1. As the storm began to transition from an ordinary, isolated cell to a supercell, the rain and hail mass fluxes rebounded and then continued to increase by an order of magnitude over the next hour despite the fact that there were no detected CG lightning flashes. Relative maxima in the peak hail rate and hail mass flux are coincident with a relative maximum in the IC flash rate at 1735 MDT as the quasi-steady supercell underwent a pulse in growth. Next, when the positive CG lightning began and the IC flash rate rapidly increased at 1750 MDT, the rain mass flux was steady at peak values (around 3 - 4 ´ 106 kg s-1) and the hail mass flux began a broad maximum (around 3 ´ 105 kg s-1) which lasted until 1830 MDT during the collapse of a high reflectivity convective core (see Figs. 8 and 11). The hail rate increased with the IC flash rate and both reached a maximum value around 1810 MDT of 35.5 mm h-1 and 16.6 flashes min-1. Of course, the IC flash rate for supercell #1 may have continued to increase since the total flash rate for supercell #1 and #2 combined eventually reached a maximum near 55 flashes min-1 around 1850 MDT (see Fig. 12a). As shown in Fig. 7, the peak hail rate began to decrease after 1810 MDT and the hail mass flux also declined significantly after 1830 MDT. At the same time, the positive CG flash rate was at its maximum for supercell #1 as seen in Fig. 6. So, the hail rate and positive CG flash rate appear anti-correlated in the later half of this lifecycle. This temporal pattern seems to suggest that enhancement of the IC lightning flash rate and the initial production of positive CG lightning was associated with an intense pulse in cell development. However, the broad peak in the positive CG flash rate appears to occur during a peak in the rain mass flux and just after a peak in the hail mass flux at the surface associated with the collapse of this pulsing cell.


Fig. 7 Temporal trend of the peak hail rate, in mm h-1, and the hail and rain mass fluxes in 104 kg s-1 for supercell #1 from 1620 - 1900 MDT.


    To explore this idea further, a time-height cross section of hail echo volume (km3) from 1740 MDT to 1855 MDT is depicted in Fig. 8. Essentially, this plot tracks the hail volume aloft at each vertical level above the freezing level using criteria 6 from Table A1. Fig. 8 also shows the evolution of the associated surface area of large hail which has fallen out of the storm (using category 3 and 4 from Table A1 at the lowest grid level z = 0.5 km AGL). First, it is important to notice the temporal consistency between the amount of hail production aloft and the surface area of large hail reaching the ground. Hail aloft inferred from multiparameter radar first appears between -10° and 0° C at 1745 MDT followed by the first large hail falling to the ground at 1750 MDT. Between 1745 and 1805 MDT, there is a surge in the amount of hail being produced aloft, accompanied by a rapid increase in the descent of large hail to the surface. The hail production aloft peaks around 1805 MDT at all levels in the mixed phase zone (0 to -40° C) with peak values of 10 - 12 km3 (between 0 and -14° C). Peak hail production aloft is quickly followed by a peak in the surface area coverage of large hail of 18.2 km2 by 1810 MDT associated with the collapsing convective cell. As expected, the amount of large hail at the surface then begins to decrease from 1810 MDT to 1835 MDT as the quantity of hail above the freezing level begins to decrease and descend in altitude. This storm wide evolution of large hail is also consistent with surface reports compiled from two chase vans that were directly in the path of the supercell (see Sec. 3.2).


Fig. 8 Time-height cross-section of the multiparameter radar inferred hail echo volume (shaded; km3) above the freezing level at each vertical level in supercell #1 from 1740-1900 MDT. Superimposed on the time-height cross-section is the temporal evolution of the areal extent of the hailswath at the ground caused by the descent of large hail (solid line with '*' marks; km2) during the same time period.


    Comparing the trends in the IC and CG lightning given in Fig. 6 to those of hail in Fig. 8, we find that the initial occurrence of large hail aloft precedes the first positive CG lightning flash by less than five minutes. The rapid increase in the IC flash rate from 1750 to 1810 MDT is coincident with the accelerated growth of hail in the entire mixed phase region and then the subsequent descent of large hail to the surface during cell collapse. Interestingly, the broad maxima in the positive CG flash rate which occurs between 1835 and 1900 MDT appears to be associated with a broad minima in both the amount of hail production aloft and the amount of large hail at the surface. Essentially, most of the large hail has already fallen out of the storm when the positive CG lightning begins its broad maximum during cell collapse. Although the first occurrence of positive CG lightning is associated with the initial production of large hail aloft, the temporal trends in the amount of large hail and positive CG lightning appear to be anti-correlated. However, both phenomena do appear to be indirectly related via the process of convective cell collapse.

    With an understanding of the temporal relationship between large hail and positive CG lightning for supercell #1, we now turn our attention to the spatial relationship between the two phenomenon. To explore this spatial correlation, a representation of the multiparameter radar inferred hail swath and the overlaid CG lightning ground strike locations is given in Fig. 9. Areas of small hail (criteria 1 and 2, Table A1) and large hail (criteria 3 and 4, Table A1) are depicted regardless of whether they are mixed with rain. Note the two small circles representing the chase vans at the following (x, y) coordinates: (-10.5 km, -15.8 km) and (-13 km, -22 km). Both vans were in the multiparameter radar inferred large hail swath and both vans reported hail from 2 - 5 cm in diameter. As can be seen in Fig. 9, the positive flash ground strike locations tended to lie outside regions of hail, especially large hail. The large majority (73%) of the flashes occurred downstream of the hail shaft, toward the northeast. This area is dominated by light to moderate rain rates (< 60 mm h-1) and is beneath the downshear anvil adjacent to the main precipitation shaft. In summary, the temporal and spatial behavior of large hail and positive CG lightning for supercell #1 appear to be anti-correlated.


Fig. 9 Depiction of the multiparameter radar inferred hail swath (both small, < 2 cm, and large, >= 2 cm, hail) from 1740 - 1900 MDT for supercell #1. Lightning ground strike locations are overlaid on the hail swath (+: positive polarity, and -: negative polarity). Location of the CSU-CHILL radar (0,0), hail chase van (-13, -22), and mobile corona point van (-10.6, -15.8) are marked by O's. Note that both vans are within the large hail streak embedded in the center of the hail swath.


    Next, we investigate the gross electrostatic structure of supercell #1 as revealed by ground based (5 m AGL) corona point measurements. In order to obtain measurements directly beneath the supercell thunderstorm, a mobile corona point sensor was placed in the direct path of the oncoming storm (reference Figs. 1 and 3 for the location relative to the storm). Point discharge measurements were obtained during a very opportune time (1748 MDT to 1848 MDT) to study trends in the surface electric field associated with hail production, enhanced IC flash rates, and a predominance of positive CG lightning. In Fig. 10, the dark trace is a plot of the 1 Hz point discharge current and the lighter trace is a smoothed version of the current (33 point moving filter). From the 1 Hz data, a copious amount of lightning transients can be seen, especially from 900 seconds (1803 MDT) and onward, which is consistent with the rapid increase in the flat plate measured total flash rate (e.g., Fig. 6). The transients are directed both upward (suggesting net negative charge aloft neutralized) and downward (suggesting net positive charge aloft neutralized). However, the strongest transients which saturated the data logger between 2400 and 3000 seconds (1828-1838 MDT) are downward directed. The smoothed corona point current gives an idea of the relative strength and polarity of the net electrostatic field aloft. There appears to be four phases in the trend of the smoothed current. From 0 to 600 seconds (1748-1758 MDT), the corona current was positive (0-2 mA) which suggests a net fair field aloft of moderate magnitude (i.e., that is net positive charge aloft). In the second phase (600-1200 seconds or 1758-1808 MDT), the negative point discharge current varied between 0 and -2 mA indicating a net foul field aloft of moderate magnitude (i.e., that is net negative charge aloft). During the third phase from 1200 to 2100 seconds (1808-1823 MDT), the current was very weakly negative, suggesting a negligible net foul field aloft (probably near the minimum E-field threshold for the on-set of corona current around ±1 kV m-1). The fourth and final phase was characterized by moderately strong positive corona current (over 3 mA) from 1823 MDT to about 1845 MDT (2100-3300 seconds) suggesting a net fair field aloft of moderate magnitude. The corona current values in all but the third phase are of similar magnitude to those beneath non-severe thunderstorms (e.g., Carey and Rutledge, 1996; Williams et al., 1989a,b).


Fig. 10  Evolution of the corona current (mA) beneath supercell #1 from 1748 - 1848 MDT.  The dark trace depicts the 1 Hz corona point current and the light trace is a smoothed (33 point running mean filter) version of the corona point current.  Each horizontal division represents five minutes.  Negative corona current indicates a foul field (or net negative charge aloft) and positive corona current indicates a fair field (or net positive charge aloft).


    To place these point discharge measurements in the context of storm precipitation structure, a time-height cross-section of reflectivity (color shaded) and LDR (contoured) directly above the corona point sensor is shown in Fig. 11 for the same times as the corona point data. As can be inferred from the values of Zh and LDR, small hail (sometimes mixed with heavy rain) fell at the corona point site from about 1808 to 1810 MDT and then from 1825 to 1834 MDT. From about 1810 to 1825 MDT, large hail (sometimes mixed with heavy rain) came down over the sensor. These radar inferred precipitation types are well supported by notes taken in the field. The time-height cross-section suggests that a combination of advection and discrete growth or dissipation occurred over the corona point. Two tilted hail shafts passed over the radar: 1) 1755-1810 MDT which produced mostly small hail and rain and, 2) 1810-1825 MDT which produced large hail. It is possible to discern that the cores were tilted toward the northeast (downshear) because the reflectivity and LDR cores are slanted with height with maximum reflectivity values occurring at higher altitudes first in time.


Fig. 11  Time-height cross-section of reflectivity (color shaded, dBZ) and the linear depolarization ratio (contoured every 2 dB starting at -28 dB) for supercell #1 directly above the mobile corona point sensor from 1748 - 1848 MDT.  Each horizontal division represents five minutes and can be directly compared to the corona point current shown in Fig. 10.


    Comparing the corona point data (Fig. 10) and radar time-height cross-section (Fig. 11), one can deduce that the arrival of the first tilted (i.e., in the downshear direction) precipitation shaft with high reflectivities aloft (i.e., the forward overhang as seen in Fig. 4) was associated with a net fair field at the surface indicating positive charge aloft. When the main precipitation shaft containing small hail and rain was overhead (around 1803 MDT), the corona point current indicates net foul field aloft (negative charge). When the second surge in reflectivity and LDR occurred associated with an increase in the amount of large hail overhead (likely through a combination of production and advection), the E-field aloft appears to be weakly foul. Finally, as the supercell moves slowly toward the ENE and the large hail falls out of the collapsing convective core, the field reverses polarity to fair values (i.e., net positive charge aloft) associated with the passage of the reflectivity gradient region beneath the initial portion of the upshear anvil. 

    With the analyses of supercell #1 complete, we now attempt to demonstrate that the temporal and spatial relationships established between electrical behavior and precipitation structure are actually valid for the entire severe storm complex observed on the afternoon of 7 June 1995.

Table of Contents



3.4. Electrical Characteristics Associated with a Severe Hail Storm Complex

    In the section above, we demonstrated that for an isolated supercell 1) the IC lightning flash rate is enhanced during vigorous storm development, 2) the positive CG lightning also begins during an intense pulse in storm development, and 3) the positive CG flash rate peaks long after most large hail has fallen out of the storm, just after a peak in the hail mass flux (associated primarily with small, D < 2 cm, hail), and during a broad peak in the rain mass flux associated with cell collapse. This behavior results in a strong anti-correlation in time and space between positive cloud-to-ground lightning and large hail for supercell #1. We now wish to determine if these relationships are valid for the entire history of this severe hailstorm complex (including supercell #2 and the merged supercell #1 and #2 at later times). First, we present (Fig. 12a) the entire lightning history of the 7 June 95 severe hailstorm complex from 1731 MDT to 2050 MDT while it was within about 60 km of the CSU-CHILL radar. The negative and positive CG lightning flash rate, the IC lightning flash rate, and the IC/CG ratio is included for all cells shown in Figs 3a-i.


Fig. 12 Temporal evolution of the lightning and precipitation rates in the 7 June 1995 storm complex depicted in Figs. 3a-i from 1731 - 2050 MDT.. a) negative and positive cloud-to-ground (CG) flash rate (# min-1), intracloud lightning flash rate (# min-1), and IC/CG ratio. b) rain mass flux (106 kg s-1), hail mass flux (105 kg s-1), and the surface coverage of large hail (km2).

    The positive CG flash rate has two broad maxima from 1836 to 1905 MDT and 1946 to 2009 MDT. There are also two secondary maxima in the positive CG flash rate at 2026 and 2038 MDT. The first broad maxima is associated primarily with supercell #1 but also has a contribution from supercell #2, especially before 1830 MDT. The highest positive CG flash rate (six minute average) is 2.7 min-1, occurs in the second maxima at 1946 MDT, and is associated with the storm complex which results from the merger between supercell #1 and #2. Although the positive CG flash rate exceeds the negative CG flash rate during the entire 3.3 hour period, the percentage of negative ground flashes increases towards the end of the period (i.e., after 2015 MDT) as the storm begins to take on the characteristics of a severe, multicell squall line.

    Before 1748 MDT when both supercell #1 and supercell #2 were in the developing phase, the ground strike polarity was predominantly negative. In addition, the IC flash rate for the cells within range was relatively small (< 5 min-1) and hence the IC/CG ratio had values between 5 and 10 which are just slightly larger then those measured for typical warm season thunderstorms (e.g., Prentice and Mackerras, 1977; Price and Rind, 1993). When both supercell #1 and #2 began to explosively develop around 1748 MDT, the predominant polarity for both storms switched to positive and the IC/CG ratio immediately ramped up to values greater than 20 as the IC flash rate increased from 5 min-1 at 1748 MDT to 55 min-1 by 1847 MDT when the flat plate became saturated by heavy rain.

    For comparison with the lightning data, Fig. 12b depicts the evolution of the surface coverage of large hail (km2), and the rain and hail mass fluxes for all cells. A comparison of Figs. 12a,b reveals that all of the positive CG lightning flash rate maxima are associated with rapid increases and subsequent peaks in both the rain and hail mass fluxes associated with collapsing cells. Associated with these descending echo cores, there were two broad maxima in the descent of large hail to the surface from 1806 to 1830 MDT and from 1905 to 1928 MDT. In addition, there were two small relative maxima at 2003 and 2021 respectively. By closely comparing the trend of the positive CG lightning (Fig. 12a) and that of large hail in Fig. 12b, it is apparent that the two phenomena are anti-correlated in time (correlation coefficient = -0.23). The two broad maxima in the surface coverage of large hail are followed by two extended peaks in the positive CG flash rate. These relationships are summarized in Table 1. Detailed analyses of the multiparameter radar data reveals that very few (< 5%) positive CG ground strike locations coincide with large hail at the surface (cf. Fig. 9).


    To further investigate the spatial relationship between positive CG lightning and storm structure, we present frequency and cumulative percentage histograms of 1) rain and hail rates which occurred at each of the 184 positive CG ground strike locations (R, H), and 2) maximum rain and hail rate which occurred within 10 km of all the positive ground strokes (Rmax, Hmax). The rain (R) and maximum rain (Rmax) rates are depicted in Fig. 13a while the hail (H) and maximum hail (Hmax) rates are depicted in Fig. 13b. Considering the widespread nature of heavy rain and hail which each separately had a peak surface area coverage of about 200 km2 in this storm complex, the results are quite striking.


Fig. 13 Frequency and cumulative percentage histograms of the precipitation rate (mm h-1) occurring at all positive ground strike locations (sample size is 184 positive CG's) and of the maximum precipitation rate within 10 km of each positive ground strike location from 1748 to 2048 MDT. a) Rain rate (R) and maximum rain rate (Rmax). b) Hail rate (H) and maximum hail rate (Hmax).

    As can be seen in Fig. 13a, the majority (54%) of positive cloud-to-ground flashes were situated in regions of light rain (<= 20 mm h-1) while 28% of flashes occurred in moderate rain (20 - 60 mm h-1), and only 18% in heavy rain (>= 60 mm h-1). By comparison, only 34% of negative cloud-to-ground lightning occurred in light rain during the same period while 46% were collocated with regions of moderate rain (not shown). The distribution for the peak rain rate within 10 km is very flat with relative maxima at 10, 100, 130, 150, and 180 mm h-1, suggesting that many positive ground strokes are within a convective scale distance (10 km) of a sometimes intense precipitation shaft.

    Similar analyses were accomplished for both hail rates (H) at the ground strike location and maximum hail rates (Hmax) within 10 km of the positive ground strike and similar results as above were obtained as shown in Fig. 13b. Only a surprising 14% of positive CG flashes struck ground in precipitation with significant hail rates (i.e., >= 2 mm h-1). On the other hand, the rather flat histogram for the maximum hail rate demonstrates that a majority (58%) of the positive CG flashes occurred within a convective scale distance from a hail shaft. Approximately 29% of the time, the hail shaft is quite intense with hail rates in excess of 20 mm h-1.

    Having demonstrated that the patterns of behavior revealed in Sec. 3.3 for an isolated supercell storm appear to apply to the storm complex as a whole, we now show that an individual cell within the multicell severe hailstorm that resulted from the merging of two supercell storms (see Sec. 3.2) also exhibited similar relationships between precipitation and lightning.

Table of Contents



3.5. An Intense Positive Cloud-to-Ground Lightning Cluster

    Observations of an intense positive CG cluster associated with a streak of large hail embedded within a larger hail swath are given in Figs. 14a,b. This cluster occurred from 1943 MDT to 2001 MDT during the period of peak positive CG flash rates for the entire storm history (c.f., Fig. 12a). During this 18 minute period, the average positive (negative) CG flash rate was about 1.5 (0.2) min-1. The positive CG peak current ranged widely from 14 kA to 103 kA with a mean value of 38 kA. Inspection of the horizontal cross-section of reflectivity (shaded) and Kdp derived rain rate (contoured, Fig. 14a) reveals that the majority of positive CG flashes struck ground adjacent to and on the left flank of the heavy precipitation (and maximum reflectivity) core centered at x = 25 km and y = 4 km. It is interesting to note that 54% of the positive flashes and 25% of the negative flashes come to ground in light rain but within 10 km of heavy precipitation, consistent with the results given in Fig. 13a. Notice the bipole pattern in the ground strike locations with negative CG flashes coincident with the heaviest rain and the center of the positive CG cluster occurring about 8 km to the NNW in light to moderate rain.


Fig. 14 Observations of an intense positive CG cluster which occurred around 1952 MDT. a) Horizontal cross-section at 0.5 km AGL of shaded reflectivity, Kdp derived rain rates contoured every 20 mm h-1, and CG ground strike locations and polarities for an 18 minute period centered on 1952 MDT. CG positions were advection corrected. b) North-south vertical cross-section of contoured reflectivity (every 10 dBZ) and shaded specific differential phase, (° km-1), through x = 28 km (see dashed line in part a). The T = 0 C and -40 C levels are indicated by dashed lines. Polarities and ground strike locations of CG flashes within 5 km either side of the depicted N-S oriented vertical plane are shown beneath the echo.


    The vertical cross-section of Zh and Kdp (Fig. 14b) through the positive CG cluster at x = 28 km depicted in Fig. 14a provides a revealing perspective of storm structure relative to CG ground strike locations. First, notice the tilt of the reflectivity contours with altitude away from the main precipitation shaft at y = 4 km. The core tilts toward the north by 4 to 10 km between the -30° C to -40° C level (6.5 to 7.7 km AGL). Also, note that the positive CG cluster occurs directly beneath this tilted core (see y = 8 - 18 km). A similar phenomena is occurring on the opposite end for the complex around y = -7 km. Observations of Kdp and Zh clearly indicate that the negative polarity CG lightning is occurring in the heaviest rain while the positive CG flashes are occurring along the edge of the heaviest precipitation and beneath the tilting core aloft similar to the findings of Rust et al. (1981a,b). Even more dramatic are the negative values of Kdp found at the -40° C level at the tops of the tilted precipitation echoes (see y = 8 - 15 km and y = -8 km).

    These specific differential phase shifts of up to -0.6° km-1 are consistent with the orientation of ice crystals in a strong electric field such that a significant component of the major axis is in the vertical, as per the findings of Caylor and Chandrasekar (1996). (See Appendix A for more details). These negative values of Kdp persisted for at least 15 minutes from 1943 to 1958 MDT (period of highest positive CG flash rate) in an irregularly shaped disk with an area of about 50 km2 almost directly above the positive CG cluster. Furthermore, the surface area of the positive CG cluster was only slightly larger than the area of the negative Kdp region aloft.

    The temporal trend of the CG lightning and precipitation field for the northern most convective cell depicted in the cross-sections of Figs. 14a,b at 1952 MDT (centered on x = 28 km and y = 5 km) is given in Fig. 15. The evolution of the positive and negative CG lightning flash rate is compared to the amount of rain, small hail, and large hail at the surface and the amount of vertically oriented ice crystals aloft. A pattern is revealed that actually repeats itself several times over in the severe hailstorms of 7 June 1995. First, note that the positive CG flash rate is increasing from 1938 - 1957 MDT while the negative CG flash rate is already decreasing toward zero. Next, the maximum number of negative CG flashes immediately follows the peak in large hail and is coincident with the maxima in rain rate and small hail at the surface, consistent with the findings of Carey and Rutledge (1996) for non-severe thunderstorms. The correlation between negative CG lightning and the area average rain rate is very high (r = 0.93). The maximum number of positive CG flashes lags the peak in large hail by 24 minutes, small hail by 19 minutes, and a pulse in heavy rain centered at x = 25 km and y = 4 km (Fig. 14a) by about 5 minutes. Apparently, positive CG lightning is one of the final acts of a severe convective cell undergoing dissipation and eventual collapse.


Fig. 15 Evolution of the surface coverage of small hail and large hail (km2), the area averaged rain rate based on Kdp (mm h-1), the volume of vertically aligned ice aloft (km3 / 10), and the number of positive and negative cloud-to-ground (CG) lightning flashes from 1929 - 2003 MDT for the northern most convective cell centered on x = 28 km and y = 5 km in Figs 14a,b. The hail echo volume aloft for this same cell is presented in Fig. 16.


    In Fig. 16, a time-height cross-section of hail echo volume (similar to Fig. 8) for this convective cell is compared to a trace of the positive CG lightning flash rate. From this depiction, it is obvious that the positive CG flash rate rapidly increases following the rapid growth and subsequent descent of hail in a collapsing convective core. By analyzing the behavior of this individual hail swath and positive CG cluster, the anti-correlation between large hail and positive CG lightning is much more significant (r = -0.85). A similar temporal anti-correlation between small hail and positive CG lightning exists as well (Fig. 15). Lastly, note that the trend in the amount of vertically aligned ice crystals aloft closely follows the trend in the positive CG lightning (r = 0.75) as seen in Fig. 15. The quantity of vertically oriented ice aloft peaks just 5 minutes prior to the peak in the positive CG flash rate. Correlation coefficients between the CG lightning and the precipitation amounts are summarized in Table 2.



Fig. 16 Time-height cross-section of the multiparameter radar inferred hail echo volume (shaded; km3) aloft for the same convective cell analyzed for surface precipitation rates in Fig. 15. The hail echo volume for this cell is presented above the freezing level at each vertical level from 1919 - 1957 MDT. Superimposed on the time-height cross-section is the number of positive cloud-to-ground lightning flashes (solid line with '+' marks) during each five minute period associated with the collapse of this hail shaft. The positive CG lightning data is repeated from Fig. 15 for the purpose of comparison.

Table of Contents



4. Discussion

4.1 Electrical and Precipitation Characteristics of a Severe Hailstorm

    Our analyses of a severe hailstorm complex have revealed some interesting relationships between the multiparameter radar inferred precipitation structure, lightning, and the surface electric field. The following points summarize the key findings of this study

        1) This severe hailstorm was characterized by an unusually large intracloud flash rate (up to 55 flashes per minute) and an enhanced IC/CG ratio (> 20). This is similar to a tornadic storm studied by MacGorman et al. (1989) and the early sferic studies reviewed by MacGorman (1993) in which the intracloud flash rate increases while cloud-to-ground lightning activity is usually decreasing or steady during the severe stage of storms. As suggested by MacGorman et al. (1989), intense updrafts in severe storms could enhance IC flash rates to the detriment of CG flashes by carrying negative charge to higher altitudes and for longer periods of time compared to ordinary thunderstorms (i.e., elevated dipole hypothesis).

       2) Measurements of the sub-storm discharge current (i.e., a useful proxy for the surface electric field) suggest that the storm was dominated by net negative charge aloft during the passage of a high reflectivity core containing heavy rain and hail. There appeared to be an inverse relationship between the strength of the net foul field (i.e., negative charge) aloft and hail size. As multiparameter radar inferred hail size aloft increased (confirmed by surface reports), the magnitude of the net foul field aloft decreased. Surface observations also suggest that there was net positive charge aloft (i.e., fair field) as both the downshear anvil and the reflectivity gradient region beneath the initial portion of the upshear anvil passed overhead. The strongest fair field was associated with the collapse of a hail shaft to the northeast as the reflectivity gradient region beneath the initial portion of the upshear anvil passed overhead. 

    Since a non-negligible amount of precipitation sized ice particles occurs in the intervening region between the corona point and the non-precipitation sized ice particles in the forward anvil, it is somewhat ambiguous as to whether the precipitation sized ice particles (roughly corresponding to Zh > 25 dBZ) in the reflectivity gradient region on the upshear side of the main precipitation core or the non-precipitating ice above this region carry predominantly positive charge. Clearly, additional in-situ particle charge and electric field measurements in this region are required to definitively answer this question.

        3) Similar to recent studies by MacGorman and Burgess (1994) and Stolzenburg (1994), this severe hailstorm complex produced a predominance (74%) of positive polarity cloud-to-ground lightning. The six minute averaged peak positive CG flash rate was 2.7 min-1 and the peak positive CG flash density was approximately 0.1 km-2 h-1. These values are comparable in magnitude but on the low side of the typical range in values for peak negative CG flash rates and densities in non-severe storms (Stolzenburg, 1990).

        4) This severe hailstorm complex produced predominately positive CG lightning despite complex changes in storm morphology (including supercell, multicell, and squall-line structures; see Sec. 3.2). During all of these different phases, positive ground strikes dominated. This is in contrast to earlier studies (Branick and Doswell, 1992; Curran and Rust, 1992) which suggested that the predominance of positive CG lightning could be a unique signature of low-precipitation supercells. As shown in this study and in MacGorman and Burgess (1994), this is not the case. Predominant positive CG lightning during the mature phase can occur in a wide variety of storm types, including classic supercell and multicell storms. Apparently, the common characteristic of most storms dominated by positive ground discharges is storm severity (e.g., the presence of large hail and possibly tornadoes).

        5) In MacGorman and Burgess (1994) and Stolzenburg (1994), a loose correlation between large hail occurrence and positive ground discharges was suggested using surface hail reports. Detailed analysis of multiparameter data in this case revealed that positive CG lightning and large hail are anti-correlated in time. The peak in the amount of large hail reaching the surface typically occurs 20 - 35 minutes prior to broad maxima in the positive CG ground flash rate. If this relationship holds for most severe hailstorms, then it has important implications for the nowcasting of severe weather using CG lightning data.

        6) Analysis of multiparameter radar data also revealed that the majority of positive CG lightning struck ground in regions of light rain with little or no hail. In general, the positive ground strike locations appeared to avoid regions of active hail fall. This is similar to the findings of Changnon (1992) who found that lightning generally struck ground to the left or right flank of hailstreaks in his study of lightning in Illinois hailstorms. Since these storms occurred in a different geographical region and likely under different meteorological settings, we are unable to generalize the results regarding hailswaths and positive CG lightning.

        7) Maxima in the positive CG lightning flash rate tended to occur during and just after the collapse of high reflectivity convective cores containing hail and heavy rain. Typically, large hail falls out of the collapsing cores first. As the cell continues to collapse, the rain and hail mass flux (associated primarily with small hail) near the surface rapidly increases. Coincident in time with this rapid increase in the surface mass flux, clusters of positive CG lightning typically occur. This behavior is similar to the findings of Seimon (1993) who studied an F5-tornado producing supercell in Illinois. In this storm, the peak positive CG flash rate occurred during a major downburst which resulted from the descent of an exceptionally large elevated reflectivity core.

        8) Observations of persistent negative Kdp shifts near the top of a tilted convective core may suggest the presence of oriented ice crystals in a strong electric field (Caylor and Chandrasekar, 1996) almost directly above the region where an intense positive CG cluster occurred. Weinheimer and Few (1987) calculated that columnar ice crystals with major dimensions between 200 mm and 1 mm would be aligned vertically in a vertically directed electric field of about 100 kV m-1. Recent research by Marshall et al. (1995) suggests that lightning may occur whenever the electric field exceeds the breakeven field. For the altitudes at which the negative Kdp signatures occurred (5.5 - 8.5 km AGL or 7 - 10 km MSL), the breakeven field ranges from 61 to 88 kV m-1. Since the E-field required to orient the ice crystals in the near vertical is on the order of the breakeven field at these altitudes, one could speculate that the positive ground discharges located almost directly below emanated from this region of moderate reflectivity near storm top. This observation is consistent with the findings of Rust et al. (1981a,b) who showed that positive CG flashes emanated from high in the storm (up to 15 km) and did not occur in the precipitation core of a severe storm. Of course, further observations such as in-situ electric field measurements, visual observations, or lightning flash mapping using acoustic, electric field change, or interferometry techniques would be necessary to substantiate this claim.

Table of Contents



4.2 Positive Cloud-to-Ground Lightning in Severe Storms: Hypotheses

    Although these observations of one severe storm provide insufficient grounds upon which to judge various hypotheses for explaining a high fraction of positive CG lightning flashes, we feel that a few speculative comments are warranted. Current hypotheses for explaining the occurrence of high positive CG flash densities in mature thunderstorms typically fall into one of two categories: 1) the tilted-dipole, and 2) the enhanced lower positive charge (sometimes called an inverted dipole).

    A tilted-dipole is a normal polarity dipole (positive over negative; Wilson, 1920) in which the positive charge carried on ice crystals is tilted downshear by storm relative winds from the precipitation shaft which carries negative charge, thus resulting in an electrical dipole tilted with respect to the vertical. As the positive charge center moves outward from directly above the negative charge center, it is no longer shielded from the earth's surface by the negative charge center. This charge arrangement could then facilitate the occurrence of positive flashes to ground. MacGorman and Nielsen (1991) and Branick and Doswell (1992) both suggested this mechanism for explaining the occurrence of positive CG lightning flashes in supercells. Some limited evidence for this mechanism can be found in our observations. For example, a majority of positive ground discharges tended to occur in light rain with little or no hail on the downshear side of a tilted precipitation shaft (see Figs. 9; 13a,b; and 14a,b). However, not all positive CG flashes exhibited this behavior. Some positive ground strokes occurred on the upshear side and a small percentage were found in heavy precipitation (see Fig. 14a). In addition, the tilted dipole hypothesis fails to elucidate why there are so few negative cloud-to-ground lightning flashes (e.g., see Table B1). If the dipole is merely tilted, one would not expect there to be a preference for positive ground discharges since the negative charge center is still assumedly closer to ground.

    MacGorman and Nielsen (1991) and Williams et al. (1991) independently proposed that the occurrence of positive ground flashes in severe hailstorm could be the result of an enhanced region of significant positive charge at low levels in thunderstorms. Observations of a lower positive charge in some thunderstorms were first reported by Simpson and Scrase (1937). Jayaratne et al. (1983) first proposed that this lower positive center could be the result of the non-inductive charging mechanism (NIC; e.g., Takahashi, 1978; Jayaratne et al., 1983; Saunders et al. 1991). Williams (1989) later supported this hypothesis with a critical review of thunderstorm scale observations. Under certain circumstances, graupel or hail particles can obtain positive charge and ice crystals acquire negative charge during a rebounding collision in the presence of supercooled cloud liquid water. The gravitational separation of these particles could result in a lower positive charge. This "charge reversal" process is the opposite of what is normally hypothesized to occur under "typical" conditions in which the graupel particle obtains negative charge and the ice crystal acquires positive charge. Charge reversal occurs at temperatures warmer than some threshold between about -10° C and -20° C (depending on liquid water content) or at some very high liquid water content, regardless of temperature.

    Williams et al. (1991) point out that most hailstones with diameters of 1 cm or larger would experience wet growth at high liquid water contents for which positive charging of the precipitation particle might occur. Based on the laboratory results of Takahashi (1978), Williams et al. (1991) and Williams (1995) suggest that the prevalence of wet growth conditions for riming hail may explain the presence of positive CG clusters in severe thunderstorms. However, the laboratory experiments of non-inductive ice-ice collision charging by Saunders and Brooks (1992) suggest that charging of the rimer (i.e., simulated hailstone) becomes negligible during wet growth conditions. The details of this controversy and their implications are further discussed in Saunders (1993 and 1995) and Williams and Zhang (1993) and Williams (1995).

    Based on their hypothesis, Williams et al. (1991) suggest that "new emphasis be placed on developing polarization-diversity radar techniques for identifying the growth regime of graupel particles." Our study represents the first polarimetric radar investigation of this question. Using enhanced values of the linear depolarization ratio (see Appendix A and Table A1), we have demonstrated that the peak occurrence of likely regions of large hail in wet growth preceded the peak positive CG lightning flash rate by 20 - 35 minutes. Given the temporal and spatial anti-correlation between large hail and positive CG lightning, it does not seem likely that the non-inductive charging of hail during wet growth and collisions with ice crystals was the source for positive ground strokes in this storm. As such, our polarimetric radar observations support the laboratory results of Saunders and Brooks (1992) and are inconsistent with the hypothesis put forward in Williams et al. (1991) for explaining the occurrence of positive CG lightning clusters beneath vigorous hailstorms.


    On the other hand, it is interesting to note that the majority of the positive CG lightning tends to occur during convective cell descent when the rain and hail mass flux are rapidly increasing at the surface. It is possible that the descent of a large number of graupel particles in a weak updraft below the so-called "charge reversal level" (e.g., Jayaratne et al., 1983) to warmer temperatures (0° C to -15° C) causes the graupel to charge positively and form an enhanced and descending positive charge layer. This enhanced lower positive charge center could easily explain the preference for positive polarity ground discharges over negative CG lightning flashes. A similar hypothesis was suggested by MacGorman and Burgess (1994). Potential evidence against this hypothesis is that the majority of positive CG flashes occurred in light rain, avoiding the areas of heaviest precipitation. In addition, the corona point sensor data suggest that the storm was typically characterized by negative charge aloft as the convective cores passed overhead and by positive charge when an anvil region was above the instrument (reference Figs. 10 and 11, 1748 - 1833 MDT). The only exception to this behavior was the fair field measured by the corona point sensor as precipitation sized ice passed overhead in the high reflectivity gradient region beneath the initial portion of the upshear anvil (reference Figs. 10 and 11, 1823 - 1833 MDT). This result suggests the need for in-situ measurements in this region to determine whether ice crystals in the upshear anvil or the intervening precipitation in the reflectivity gradient region carry the predominately positive charge.

    The separation of charge as air bubbles burst at the surfaces of a large number of melting, millimeter sized graupel or small hail particles could be an alternative (or additional) source of the hypothesized enhanced lower positive charge center (Dinger and Gunn, 1946; Drake, 1968). According to the laboratory experiments of Drake (1968), strong convection currents developed in the melt water of millimeter (1 - 5 mm) sized ice particles. Bubbles released from the ice during melting were swept up by the currents and rapidly transferred to the surface where they burst. The bursting of these bubbles in melt water was accompanied by a significant separation of electric charge, leaving the melt water surrounding the ice positively charged. It is interesting to note that the peak positive CG lightning flash rates occurred almost simultaneously with rapid increases in the rain mass flux (e.g., Figs. 7, 12, and 15), associated with the melting of large numbers of millimeter sized graupel and hail particles as confirmed by multiparameter radar data (not shown). Also, surface and multiparameter radar observations suggest that moderate-to-heavy rain and pea sized hail were falling over the corona point sensor associated with cell collapse when the peak positive electric field was recorded at the surface around 1833 MDT (Fig. 11). These observations are generally consistent with the generation of an enhanced, lower positive charge center through the Drake (1968) melting mechanism.

    Based on our limited observations, we can only speculate on the possibility of a third mechanism. Since peaks in the positive CG flash rate appear to be coincident with the collapse of intense precipitation cores, perhaps it is the unshielding of a large reservoir of upper level positive charge by the action of rapid precipitation echo descent which results in a favorable electrostatic condition for positive CG lightning. Once much of the heaviest precipitation and associated charge has fallen out of a convective cell, the upper level positive charge typically resident on ice crystals would no longer be shielded from ground by the lower negative charge typically resident on precipitation particles. The longevity and intensity of severe storms might allow an extremely large reservoir of positive charge to be accumulated aloft. Assuming that the enhanced amount of precipitation mass descending to the surface carries significant negative charge, this mechanism could also explain the predominance of positive over negative CG lightning flashes. We speculate that the precipitation current (Jp) in some severe storms may be augmented to the point of reducing the electrical requirement for other current sinks such as the lightning current (Jl) associated with negative ground discharges.

    This "precipitation unshielding" hypothesis is broadly consistent with all of our observations: 1) positive CG lightning is associated with the descent of an intense convective echo, 2) large hail falls out of the storm ten's of minutes before the peak in positive CG lightning, 3) the rain and hail mass flux (associated with mostly small hail) is either rapidly increasing or experiencing maxima during the peak in positive CG lightning activity, 3) many of the positive ground strokes occur in light rain and little or no hail, 4) the corona point data suggest that the gross electrostatic structure of the storm could be characterized by a normal polarity dipole (negative charge aloft as the convective cores passed by and positive charge aloft as either an anvil passed overhead or after the descent of a convective core, and 5) radar observations suggest that ice crystals aloft were oriented in the near-vertical by a strong electric field on the order of the breakeven field in a large and persistent region which was almost directly above a positive CG lightning cluster.

    Clearly, more case studies of positive CG producing severe storms are required before any definitive statement can be made regarding causal mechanisms. In particular, more polarimetric radar studies would be useful since they can remotely probe cloud precipitation in all regions of these severe storms. Of course, the most powerful approach would be to study several of these storms using a host of observational techniques in concert, including 1) polarimetric radar observations set in the context of the three dimensional wind field obtained by dual-Doppler synthesis, 2) in-situ (e.g., aircraft) microphysical, electric field, and particle charge measurements, and 3) a detailed lightning observational network (NLDN, flat plate antennas, surface field mills, and interferometer).

    Future studies of the unique electrical aspects of severe storms will benefit both the atmospheric electricity community by giving us a better understanding of thunderstorm electrification mechanisms and the operational forecasting community by further exploring the potential of lightning data in the identification and nowcasting of severe weather (e.g., Knapp, 1994).

Table of Contents



Appendix A: Multiparameter Radar: Analysis Method

    Using the gridded polarimetric radar observables (Zh, Zdr, Kdp, ρdp, and LDR), it is possible to estimate first the predominant precipitation type and then the amount at each grid point in the storm. In this study of a severe hailstorm, we are particularly interested in two objectives: 1) determine the bulk hydrometeor type with emphasis given to the identification of small (< 2 cm) and large (>= 2 cm) hail, and 2) calculate rain and hail rates near the surface throughout the storm.

    Jameson (1985a) was one of the first investigators to suggest that polarimetric radar measurements could provide a possible framework for precipitation identification. In particular, he discussed the discrimination of liquid from solid hydrometeors and discrimination among solid hydrometeors. More recently, Doviak and Zrnic´ (1993) proposed a multiparameter radar decision process that partitions the five-dimensional space of Zh, Zdr, Kdp, ρdp, and LDR so that each category uniquely corresponds to a distinct hydrometeor type. This method of bulk hydrometeor identification is based on both modeling and observational studies. For detailed reviews of these studies and a general discussion of the technique, see Jameson and Johnson (1990) or Doviak and Zrnic´ (1993). 

    These techniques have been applied in a variety of case studies. For example, the method was used successfully in a severe hailstorm by Zrnic´ et al. (1993) to identify various hydrometeor types including large hail, small hail, rain, mixed phase precipitation (e.g., rain and hail mixture), graupel, and ice crystals. Höller et al. (1994) employed C-band polarimetric observations of LDR and Zdr to identify rain, graupel, hail, and mixed phase precipitation in the investigation of a hybrid-type hailstorm. Carey and Rutledge (1996) utilized bulk hydrometeor identification (without LDR) to differentiate between rain and graupel in a multicell storm in CO in order to correlate the evolution of the multiparameter radar inferred graupel volume to that of lightning and storm electric fields. Jameson et al. (1996), Ramachandran et al. (1996), and French et al. (1996) used multiparameter radar observations including Zh, Zdr, and LDR of isolated Florida thunderstorms during the CaPE (Convective and Precipitation/Electrification) project to demonstrate that initial electrification was coincident with the freezing of supercooled raindrops above the -7° C level. In this study, we utilize similar methods to track the production of hail, especially large hail, in a positive CG producing storm in order to assess the temporal and spatial relationship between the two phenomenon.

    A summary of the bulk hydrometeor method utilized in this study is shown in Table A1. The range of each multiparameter radar observable is an adaptation of the proposed values found in Table 8.1 of Doviak and Zrnic´ (1993). A few minor adjustments in these ranges were made in order to 1) account for the effect of imperfect measurements, 2) incorporate more recent modeling and observational results, and 3) maximize agreement between surface reports of precipitation type and size from the public (Storm Data, 1995) and two storm chase vans as summarized in Sec. 3.2. The primary goal of the analysis below the melting level (2.4 km AGL, cf. Fig. 2) was to locate regions of both small and large hail (categories 1 and 3 in Table A1), regardless of whether the hailstones were mixed with rain as they often are (categories 2 and 4). 

    An elevated value of Kdp is the best indication of rain since it is linearly proportional to the mass-weighted mean axis ratio, as discussed in Jameson (1985b). Typically, Kdp >= 0.5° km-1 suggests the presence of significant rain (R > 22 mm h-1, see Eqn. A3 in Table A2). Note that Zdr can be an unreliable measure of the presence of rain in mixed phase precipitation if hail dominates the reflectivity. Differentiating mixed phase precipitation from pure rain depends largely on the magnitudes of LDR and ρdp. Assuming Kdp suggests the presence of rain, the probability of hail being present increases with decreasing Zdr, decreasing ρdp, and increasing values of LDR (Bringi et al., 1986; Aydin and Zhao, 1990; Balakrishnan and Zrnic´, 1990a; Zrnic´ et al., 1993). Similarly, the determination of hail size (small vs. large) also falls largely on the values of LDR and ρdp. Increasing values of LDR and decreasing values of ρdp are associated with increasing probability of the presence of large hail (Bringi et al., 1986; Aydin and Zhao, 1990; Balakrishnan and Zrnic´, 1990a; Zrnic´ et al., 1993). The presence of large hail is often associated with the occurrence of negative Zdr (Balakrishnan and Zrnic´, 1990a). However, since this is not always the case (e.g., tumbling or spherical hailstone between 2 and 4 cm in diameter), it is not required.


    Above the freezing level, our intention is to locate probable locations of hail (category 6 in Table A1) which are often associated with an enhanced value of LDR (Bringi et al., 1986) due to hail's increasing oblateness with size (Knight, 1986), tendency to tumble (Knight and Knight, 1970), and high density or dielectric factor (especially if it is experiencing wet growth). The chosen threshold for LDR (> -26 dB) is based on 1) available modeling studies (e.g., Bringi et al., 1986; Aydin and Zhao, 1990), 2) the general agreement between analyzed areas of LDR > -26 dB with regions of hail growth in conceptual models of supercell thunderstorms (e.g., hail cascade along the edge of the weak echo region in Browning and Foote, 1976; reference Figs. 4 and 5a,b), and 3) consistency between the temporal and spatial trends in the identification of hail above and below the melting level (see Fig. 8). Of course, it is possible that a few nearly spherical hailstones with no surface irregularities could be characterized by LDR <= -26 dB. We therefore suggest that our estimates of hail volume aloft represent a lower bound since no other hydrometeor type in areas of uncorrupted data could be characterized by the ranges of multiparameter variables listed in Table A1.

    The final hydrometeor category of interest in Table A1 is that of vertically oriented ice crystals (category 7). Caylor and Chandrasekar (1996) demonstrated that negative values of Kdp aloft (7 to 11 km MSL) are associated with vertically oriented ice crystals in a strong electric field. By "vertically oriented," we strictly mean that the component of the major axis in the vertical exceeds 50%. In fact, an ice crystals only needs to exhibit orientation of the major axis more than 45° from the horizontal polarization plane (which is parallel to the ground for an elevation angle of 0°) to result in a negative value of Kdp. The closer the orientation angle is to 90° from the horizontal polarization plane; the higher Kdp will be for a fixed concentration of ice crystals (Caylor and Chandrasekar, 1996). For a fixed orientation angle, Kdp will increase with increasing numbers of ice crystals. Vivekanandan et al. (1994) demonstrated that Kdp scales with the No parameter of an assumed exponential distribution for ice crystals. Similar observations of oriented ice crystals in circular polarization radar data were first found by Hendry and McCormick (1976) and have been investigated more recently by Metcalf (1995,1997) and Krehbiel et al. (1996). Weinheimer and Few (1987) calculated that columnar ice crystals with major dimensions between 200 mm and 1 mm would be aligned with a vertical electric field of about 100 kV m-1. In this study, "vertically" oriented ice crystals are associated with spatially coherent areas of Kdp < -0.25 km in order to differentiate from noise (Bringi et al., 1996) and Zh < 40 dBZ in order to differentiate from vertically oriented hail (Doviak and Zrnic´, 1993) at temperature less than -20° C (altitudes > 5.1 km AGL or 6.6 km MSL). Small conical graupel oriented in the vertical with reflectivities less than 40 dBZ will typically not have a sufficiently large dielectric constant (due to reduced ice density) or oblateness to cause measurable Kdp.

    With regions of rain, hail, and mixed phase precipitation identified, the next analysis step was to utilize the available multiparameter radar data to calculate the rain and hail rate at each grid point near the surface (0.5 km AGL) using Eqns. A1 - A4 summarized in Table A2. Light rain (<= 20 mm h-1) is still measured best with a simple R-Zh relationship (Eqn. A1), as discussed in Chandrasekar and Bringi (1988). Chandrasekar et al. (1990) demonstrated that precipitation estimators using a combination of Zh and Zdr (Eqn. A2) provide superior estimates at intermediate rain rates (20 - 60 mm h-1) and that equations based on Kdp (Eqn. A3) perform best in heavy rain (>= 60 mm h-1). In order to avoid the possibility of hail contamination in R(Zh, Zdr), Zdr was corrected when the ice fraction exceeded 10% using the Zh, Zdp method of Golestani et al. (1989). The difference reflectivity, or Zdp, is defined as Zdp = 10 log (Zh - Zv) [dB]. The relationship between Zdp and Zh for rain is linear (Golestani et al., 1989). As a result, it is possible to estimate the fractional contribution of ice to the measured reflectivity by calculating the reflectivity deviation from the Zh, Zdp line. Using the Zh/Zdp line derived in Carey and Rutledge (1996) (their Eqn. 5) from data in a Colorado thunderstorm, the ice fraction was calculated at each grid point. If the ice fraction of the total reflectivity exceeded 10%, it was subtracted from Zh and then Zdr was re-calculated before estimating the rain rate, R(Zh, Zdr), thus minimizing contamination from hail.

    Balakrishnan and Zrnic´ (1990b) demonstrated that the hail rate could be calculated by using Kdp and Zh to separate the hail contribution to the reflectivity factor, Zhail = Zh - Zrain(Kdp) [mm6 m-3]. The reflectivity from rain is estimated by combining the Kdp, R relationship in Eqn. A3 and Zh, R relationship in Eqn. A1 to obtain Zrain(Kdp) = 77899 (Kdp)1.186. According to Balakrishnan and Zrnic´ (1990b), the value of Zhail will possess sufficient precision to use if it is within 7 dB of Zrain. If this constraint was met, then the hail rate in mm h-1 (liquid equivalent) was calculated by substituting Zhail into Eqn. (A4) which is based on the Cheng and English (1983) hail size distribution. Otherwise, it was assumed that the hail rate was negligible. Since there are several assumptions which must be made in order to estimate the hail rate (e.g., a Zh-R relationship; a hail size distribution; spherical or tumbling hail), there could be occasional point-to-point uncertainties in the hail estimate as conditions vary from our assumptions. However, the analyses in this study depend more on storm wide spatial and temporal trends rather than point-to-point estimates of precipitation rate. Therefore, we interpret our results with confidence.

Table of Contents



Appendix BCloud-to-Ground Lightning: Statistics and Assessment of the Validity of Positive Flashes

    To demonstrate that the positive cloud-to-ground lightning flashes analyzed in this study are probably not part of a subset of newly detected discharges by the NLDN which may or may not be positive CG lightning, we first analyze cloud-to-ground lightning statistics. Negative CG lightning data is included for reference. In Fig. B1, we present frequency and cumulative percentage histograms of peak current (kA) for all cloud-to-ground flashes produced by the severe hailstorm complex under study. The statistics associated with these peak currents are given in Table B1. Negative CG lightning flashes have a median peak current of -14.4 kA with values in a rather tight range from -5.3 kA down to -45.2 kA. Interestingly, about 80% of all negative ground flashes are characterized by peak currents less than 20 kA. In contrast, positive CG peak currents have a median value of 32.7 kA and range from a minimum of 13.7 kA to a maximum of 110.6 kA. As seen in Fig. B1, approximately 73% of the positive CG peak currents are between 20 kA and 50 kA. About 20% of the flashes occur in a long tail from 50 kA up to 120 kA with the remaining flashes (7%) occurring between 10 kA and 20 kA.


Fig. B1 Frequency and cumulative percentage histograms of the positive and negative CG peak currents (kA) for all ground flashes under study.



    As discussed in Sec. 2, the NLDN underwent changes in sensitivity and discrimination criteria between the 1994 and 1995 warm season (Cummins et al., 1995) which resulted in the inclusion of a previously undetected population of small amplitude positive discharges. At this time, it is unclear whether these event are long vertical cloud discharges or a new population of previously undetected positive cloud-to-ground flashes. However, recent research suggests that these newly detected events exhibit low currents from 5 kA to 15 kA (K. Cummins, private communication). As can be clearly seen from the histogram in Fig. B1, the overwhelming majority of flashes in this storm are not part of this newly identified population. As a matter of fact, only 3 out of 184 (1.6%) positive CG flashes under study for this severe hailstorm have peak return stroke currents of under 15 kA.

    In order to pursue this matter further, we have analyzed available corona point sensor data and flat plate data. Although it is not generally possible to unequivocally differentiate IC flashes from positive CG flashes with the data from either of these devices alone (e.g., Beasley, 1985), it is possible to demonstrate that all available data is consistent with the NLDN identification of positive CG lightning. We obtained approximately 1.3 hours (1730 - 1850 MDT) of quality flat plate data while the severe hailstorm complex was within range (40 km). During this period, there were 40 NLDN identified positive CG flashes within 40 km of the flat plate which was located at the CSU-CHILL (see Fig. 1 and Fig. 3). Although the storm complex was in range for a longer period of time, the flat plate data was later (after 1850 MDT) contaminated by heavy rain which eventually saturated the system's electronics.

    Our procedure was to analyze the flat plate waveform for all NLDN detected positive CG flashes within 5 km to 25 km of the CSU-CHILL. We chose to limit the range to outside of 5 km in order to minimize the chance for IC flashes being mistakenly identified as positive CG flashes (Brook et al., 1989). We also chose to restrict the range to be within 25 km so that the waveform recorded by our flat plate would have sufficient amplitude to be above the noise level of the instrument (the detection efficiency drops rapidly for ranges greater than 35 km). These range requirements decreased our sample of NLDN detected positive CG lightning flashes to thirty with peak currents ranging from 15.6 kA to 85 kA. All 30 of the analyzed waveforms had a large amplitude, initially upward going (positive deflection) electrostatic field change. A positive deflection represents a positive electric field change and, in our convention adapted from Brook et al. (1989), represents a positive stroke. Typically, this upward going deflection was followed by a downward going (negative deflection) ionospheric reflection. For about 30% of the analyzed flashes, the subsequent downward going ionospheric reflection was evident but barely discernible. A typical example representing the other 70% of the waveforms which had more obviously classic positive CG lightning characteristics is shown in Fig. B2. This flash occurred at approximately 1839:13 MDT and was detected by the NLDN with a peak current of 30.8 kA (close to the median positive CG flash in Table B1) at about 11.7 km due south of the CSU-CHILL radar. As in most of the 30 positive CG flashes, the initial upward going deflection temporarily saturated the flat plate electronics since the instrument was set at the second highest gain setting during the analysis period.


Fig. B2 Filtered output of the flat plate antenna at 1839:13 MDT (0039:13 UTC) on 7 June 1995, corresponding to an NLDN identified positive CG lightning flash with a peak current of 30.8 kA.

    By comparing the corona point data shown in Fig. 10 to seventeen NLDN detected positive CG flashes located within 10 km of the mobile corona point sensor located to the southwest of the CHILL (see Fig. 1 and Fig. 3) from 1748 MDT to 1859 MDT, we were able to further support the NLDN polarity identification. All seventeen of the NLDN identified positive CG flashes within 10 km of the corona point were characterized by large amplitude, downward directed transients which started and recovered to "fair" weather (positive) field as the storm was overhead. Two of the strongest downward transients associated with positive cloud-to-ground lightning are located between 2400 and 3000 seconds in Fig. 10. These transients occurred while the corona point measurements indicate net positive charge aloft. The fact that the transients were downward directed and the current then recovered within seconds to positive values suggests that net positive charge aloft was neutralized, as in a positive CG lightning flash. Of course, it also possible that an IC lightning flash could generate a similar transient (notice the other two strong downward deflections between 2000 and 3000 seconds in Fig. 10 that were not detected by the NLDN as positive CG lightning flashes). However, the strong support for the NLDN positive CG identifications comes from the consistent pattern of downward directed transients for all 17 NLDN detected positive CG flashes within 10 km of the corona point sensor.

    Based on the above analyses, we feel that most if not all of the 184 positive CG lightning flashes detected by the NLDN in this severe hailstorm complex were correctly identified because 1) 98.4% of the positive CG flashes have peak currents above the suspected range of 5-15 kA, 2) all available flat plate waveforms associated with the NLDN positive CG lightning flashes within 5 km to 25 km of the CSU-CHILL were similar and consistent with known positive CG waveform at a distance, and 3) all NLDN positive CG flashes within 10 km of the mobile corona point were accompanied by transients in the point discharge current which suggest net positive charge being neutralized aloft.

Table of Contents




    We thank the entire CSU-CHILL staff for their dedicated efforts during the 1995 field season. We acknowledge Pat Kennedy's (CSU-CHILL Manager) and Dave Brunkow's (CSU-CHILL Chief Engineer) assistance with radar data calibration and quality control issues. Dave Brunkow, Bob Bowie, and Ken Pattison assisted with the installation of the flat plate antenna and the outfitting of the mobile corona point van. We thank Prof. V. N. Bringi, Dr. John Hubbert, and Mr. Pat Kennedy for useful discussions regarding the identification of large hail with multiparameter radar data and the correction of attenuation and differential attenuation in wet hail. Constructive comments from two anonymous reviewers greatly improved the clarity of the text. The software used to filter differential phase and calculate specific differential phase was graciously supplied by V. N. Bringi. Surface hail reports from the "hail chase van" were provided by John Hubbert. Rain gauge data at the CSU-CHILL radar was obtained from Dr. John Beaver. We acknowledge Jon Erdman, Tim Lang, and Rick Lucci for their assistance with manning and maintaining the various cloud electricity instruments and for forecasting support during the entire 1995 field season. We also thank Dr. Walt Petersen, Dr. Terry Schuur, Jon Erdman, Rick Lucci, and Dr. Donald R. MacGorman for stimulating discussions of the electrical, microphysical, and kinematic characteristics of severe hailstorms. We acknowledge the National Aeronautics and Space Administration, Marshall Space Flight Center (NASA/MSFC) for providing the NLDN CG lightning data and Bob Rilling of the NCAR Atmospheric Technology Division (ATD) for supplying the WSR-88D RADAR data used in this study. This research was supported by the National Science Foundation (NSF) under grant ATM-9321361 and the CSU-CHILL cooperative agreement ATM-9500108.

Table of Contents




Aydin, K., and Y. Zhao, A computational study of polarimetric radar observables in hail, IEEE Trans. on GeoSci. and Remote Sensing, 28, 412-422, 1990.

Balakrishnan, N., and D. S. Zrnic*, Use of polarization to characterize and discriminate large hail, J. Atmos. Sci., 47, 1525 - 1540, 1990a.

Balakrishnan, N., and D. S. Zrnic*, Estimation of rain and hail rates in mixed-phase precipitation, J. Atmos. Sci., 47, 565-583, 1990b.

Beasley, W., Positive cloud-to-ground lightning observations, J. Geophys. Res., 90, 6131- 6138, 1985.

Branick, M. L., and C. A. Doswell III, An observation of the relationship between supercell structure and lightning ground strike polarity, Wea. Forecasting, 7, 143- 149, 1992.

Bringi, V. N., J. Vivekanandan, and J. D. Tuttle, Multiparameter radar measurements in Colorado convective storms. Part II: Hail detection studies, J. Atmos. Sci., 43, 2564- 2577, 1986.

Bringi, V. N., A. Hendry, Technology of polarization diversity radars for meteorology, in Radar in Meteorology: Battan Memorial and 40th Anniversary Radar Meteorology Conference, ed. David Atlas, Amer. Meteor. Soc., Boston, MA, 1990.

Bringi, V. N., L. Liu, P. C. Kennedy, V. Chandrasekar, and S. A. Rutledge, Dual Multiparameter radar observations of intense convective storms: The 24 June 1992 case study, Meteorol. Atmos. Phys., 59, 3-32, 1996.

Brook, M., R. W. Henderson, and R. B. Pyle, Positive lightning strokes to ground, J. Geophys. Res., 94, 13,295-13,303, 1989.

Browning, K. A., and G. B. Foote, Airflow and hail growth in supercell storms and some implications for hail suppression. Quart. J. Roy. Meteor. Soc., 102, 499-533, 1976.

Carey, L. D., and S. A. Rutledge, A multiparameter radar case study of the microphysical and kinematic evolution of a lightning producing storm, Meteorol. Atmos. Phys., 59, 33-64, 1996.

Caylor, I. J., and V. Chandrasekar, Time-varying ice crystal orientation in thunderstorms observed with multiparameter radar, IEEE Trans. on Geosciences and Remote Sensing, 34, 847-858, 1996.

Chandrasekar, V., and V. N. Bringi, Error structure of multiparameter radar and surface measurements of rainfall, Part 1: Differential reflectivity, J. Atmos. Oceanic Technol., 5, 783-795, 1988.

Chandrasekar, V., V. N. Bringi, N. Balakrishnan, and D. S. Zrnic*, Error structure of multiparameter radar and surface measurements of rainfall. Part III: Specific differential phase. J. Atmos. Oceanic Technol., 7, 621-629, 1990.

Changnon, S. A., Temporal and spatial relations between hail and lightning, J. Appl. Meteor., 31, 587-604, 1992.

Cheng, L., and M. English, A relationship between hailstone concentration and size, J. Atmos. Sci., 40, 204-213, 1983.

Cummins, K. L., E. A. Bardo, W. L. Hiscox, R. B. Pyle, and A. E. Pifer, NLDN95: A combined technology upgrade of the U.S. national network, Presented at the Intl. Aerospace & Ground Conference on Lightning and Static Electricity, Williamsburg, VA, September, 1995.

Cummins, K. L., E. A. Bardo, W. L. Hiscox, R. B. Pyle, and A. E. Pifer, A combined TOA/MDF technology upgrade of the U.S. national lightning detection network, in Proc. 10th Intl. Conf. Atmos. Electricity, Osaka, Japan, pp. 288-292, 1996.

Curran, E. B., and W. D. Rust, Positive ground flashes produced by low-precipitation thunderstorms in Oklahoma on 26 April 1984, Mon. Wea. Rev., 120, 544-553, 1992.

Dinger, J. E., and R. Gunn, Electrical effects associated with a change of state of water, Terr. Mag. and Atmos. Elec., 51, 477 - 494.

Doswell, C. A. III, Synoptic-scale environments associated with high plains severe thunderstorms, Bull. Amer. Soc., 61, 1388-1400, 1980.

Doviak, R. J., and D. S. Zrnic*, Doppler Radar and Weather Observations, 2nd Ed., 562 pp., Academic Press, San Diego, CA, 1993.

Drake, J. C., Electrification accompanying the melting of ice particles. Q. J. R. Meteorol. Soc., 94, 176 - 191, 1968.

Engholm, C. D., E. R. Williams, and R. M. Dole, Meteorological and electrical conditions associated with positive cloud-to-ground lightning, Mon. Wea. Rev., 118, 470-487, 1990.

French, J. R., J. H. Helsdon, A. G. Detwiler, and P. L. Smith, Microphysical and electrical evolution of a Florida thunderstorm, 1, observations, J. Geophys. Res., 101, 18,961- 18,978, 1996.

Fuquay, D. M., Positive cloud-to-ground lightning in summer thunderstorms, J. Geophys. Res., 87, 7131-7140, 1982.

Golestani, Y., V. Chandrasekar, and V. N. Bringi, Intercomparison of multiparameter radar measurements. Proc. 24th Conf. Radar Meteorology, Tallahassee, FL, pp. 309-314, Amer. Meteorol. Soc., Boston, Mass., 1989.

Hendry, A., and G. C. McCormick, Radar observations of the alignment of precipitation particles by electrostatic fields in thunderstorms. J. Geophys, Res., 81, 5353-5357.

Holle, R. L., and R. E. Lopez, Overview of real-time lightning detection systems and their meteorological uses. NOAA Technical Memorandum, National Severe Storms Laboratory, Norman, OK, ERL NSSL-102, 68 pp, 1993.

Höller, H., V. N. Bringi, J. Hubbert, M. Hagen, and P. F. Meischner, Life cycle and precipitation formation in a hybrid-type hailstorm revealed by polarimetric and Doppler radar measurements, J. Atmos. Sci., 51, 2500-2522, 1994.

Houze, R. A., Jr., Cloud Dynamics, 573 pp., Academic Press, San Diego, CA, 1993.

Hubbert, J., V. Chandrasekar, V. N. Bringi, P. Meischner, Processing and interpretation of coherent dual-polarized radar measurements, J. Atmos. Oceanic Technol., 10, 155- 164, 1993.

Hubbert, J., and V. N. Bringi, An iterative filtering technique for the analysis of copolar differential phase and dual-frequency radar measurements, J. Atmos. Oceanic Technol., 12, 643-648, 1995.

Hubbert, J. H., V. N. Bringi, L. D. Carey, S. Bolen, CSU-CHILL polarimetric radar measurements from a severe hailstorm in eastern Colorado, J. Appl. Meteor., accepted, 1997.

Jameson, A. R., Deducing the microphysical character of precipitation from multiple- parameter radar polarization measurements, J. Climate and Appl. Meteor., 24, 1037- 1047, 1985a.

Jameson, A. R., Microphysical interpretation of multiparameter radar measurements in rain. Part III: interpretation and measurement of propagation differential phase shift between orthogonal linear polarizations, J. Atmos. Sci, 42, 607-614, 1985b.

Jameson, A. R., and D. B. Johnson, Cloud microphysics and radar, in Radar in Meteorology: Battan Memorial and 40th Anniversary Radar Meteorology Conference, ed. David Atlas, Amer. Meteor. Soc., Boston, MA, 1990.

Jameson, A. R., M. J. Murphy, and E. P. Krider, Multiple-Parameter radar observations of isolated Florida thunderstorms during the onset of electrification, J. Appl. Meteor., 35, 343-354, 1996.

Jayaratne, E. R., C. P. R. Saunders, and J. Hallet, Laboratory studies of the charging of soft hail during ice crystal interactions, Q. J. R. Meteorol. Soc., 109, 609-630, 1983.

Jones, D. M. A., 3 cm and 10 cm wavelength radiation backscatter from rain. Proc. Fifth Wea. Radar Conf., Amer. Meteor. Soc., Boston, MA, 281-285, 1955.

Knapp, D. I., Using cloud-to-ground lightning data to identify tornadic thunderstorm signatures and nowcast severe weather, National. Weather Digest, 19, 35-42, 1994.

Knight, C. A., and N. C. Knight, The falling behavior of hailstones, J. Atmos. Sci., 27, 672-681, 1970.

Knight, N. C., Hailstone shape factor and its relation to radar interpretation of hail, J. Clim. and Appl. Meteorol., 25, 1956-1958, 1986.

Krehbiel, P. T. Chen, S. McCrary, W. Rison, G. Gray, and M. Brook, The use of dual channel circular-polarization radar observations for remotely sensing storm electrification, Meteorol. Atmos. Phys., 59, 65-82, 1996.

Lucci, R. M., Doppler radar investigation of tornadic and lightning producing storms in northeast Colorado, M. S. Thesis, Atmos. Science Paper 612, Colorado State University, 151 pp., 1996.

MacGorman, D. R., D. W. Burgess, V. Mazur, W. D. Rust, W. L. Taylor, and B. C. Johnson, Lightning rates relative to tornadic storm evolution on 22 May 1981, J. Atmos. Sci., 46, 221-250, 1989.

MacGorman, D. R., and K. E. Nielsen, Cloud-to-ground lightning in a tornadic storm on 8 May 1986, Mon. Wea. Rev., 119, 1557-1574, 1991.

MacGorman, D. R., Lightning in tornadic storms: A review, in The Tornado: Its Structure, Dynamics, Prediction, and Hazards, pp. 173-182, Amer. Geophys. Union Monograph 79, 1993.

MacGorman, D. R., and D. W. Burgess, Positive cloud-to-ground lightning in tornadic storms and hailstorms, Mon. Wea. Rev., 122, 1671-1697, 1994.

Metcalf, J. I., Radar observations of changing orientations of hydrometeors in thunderstorms, J. Appl. Meteor., 34, 757-772, 1995.

Metcalf, J. I., Temporal and spatial variations of hydrometeor orientations in thunderstorms, J. Appl. Meteor., 36, 315 - 321., 1997.

Marshall, T. C., M. P. McCarthy, and W. D. Rust, Electric field magnitudes and lightning initiation in thunderstorms, J. Geophys. Res., 100, 7097-7104, 1995.

Mueller, E. A., S. A. Rutledge, V. N. Bringi, D. Brunkow, P. C. Kennedy, K. Pattison, R. Bowie, and V. Chandrasekar, CSU-CHILL radar upgrades, in Preprints, 27th Conf. Radar Meteor., Vail, CO, pp. 703-706, Amer. Meteorol. Soc., Boston, Mass., 1995.

Orville, R. E., R. A. Weisman, R. B. Pyle, R. W. Henderson, and R. E. Orville, Jr., Cloud- to-ground lightning flash characteristics from June 1984 through May 1985, J. Geophys. Res., 92, 5640-5644, 1987.

Orville, R. E., Calibration of a magnetic direction finding network using measured triggered lightning return stroke peak currents, J. Geophys. Res., 96, 17,135-17,142, 1991.

Orville, R. E., Cloud-to-ground lightning flash characteristics in the contiguous United States: 1989-1991, J. Geophys. Res., 99, 10,833-10,841, 1994.

Orville, R. E., and A. C. Silver, Lightning ground flash density in the contiguous United States: 1992-1995, Mon. Wea. Rev., 125, 631-638, 1997.

Prentice, S. A., and D. Mackerras, The ratio of cloud to cloud-to-ground lightning flashes in thunderstorms, J. Appl. Meteor., 16, 545-550, 1977.

Price, C., and D. Rind, What determines the cloud-to-ground lightning fraction in thunderstorms?, Geophys. Res. Letters., 20, 463-466, 1993.

Ramachandran, R., A. Detwiler, J. Helsdon Jr., P. L. Smith, and V. N. Bringi, Precipitation development and electrification in Florida thunderstorm cells during Convection and Precipitation/Electrification Project, J. Geophys. Res., 101, 1599- 1620, 1996.

Reap, R. M., and D. R. MacGorman, Cloud-to-ground lightning: Climatological characteristics and relationships to model fields, radar observations, and severe local storms, Mon. Wea. Rev., 117, 518-535, 1989.

Rogers, R. R., A review of multiparameter radar observations of precipitation, Radio Science, 19, 23 - 36, 1984.

Rust, W. D., W. L. Taylor, D. R. MacGorman, and R. T. Arnold, Research on electrical properties of severe thunderstorms on the Great Plains, Bull. Amer. Meteor. Soc., 62, 1286-1293, 1981a.

Rust, W. D., D. R. MacGorman, and R. T. Arnold, Positive cloud-to-ground lightning flashes in severe storms, Geophys. Res. Lett., 8, 791-794, 1981b.

Rust, W. D., D. R. MacGorman, and S. J. Goodman, Unusual positive cloud-to-ground lightning in Oklahoma storms on 13 May 1983, in Preprints, 14th Conf. on Severe Local Storms, Indianapolis, IN, Amer. Meteor. Soc., 372-375, 1985

Rutledge, S. A., and D. R. MacGorman, Cloud-to-ground lightning activity in the 10-11 June 1985 mesoscale convective system observed during the Oklahoma-Kansas PRE- STORM Project, Mon. Wea. Rev., 116, 1393-1408, 1988.

Rutledge, S. A., C. Lu, and D. R. MacGorman, Positive cloud-to-ground lightning in mesoscale convective systems, J. Atmos. Sci., 47, 2085-2100, 1990.

Sachidananda, M., and D. S. Zrnic*, Rain rate estimates from differential polarization measurements. J. Atmos. Oceanic. Tech., 4, 588-598, 1987.

Saunders, C. P. R., W. D. Keith and R. P. Mitzeva, The effect of liquid water on thunderstorm charging, J. Geophys. Res., 96, 11007-11017, 1991.

Saunders, C. P. R., and I. M. Brooks, The effects of high liquid water content on thunderstorm charging, J. Geophys. Res., 97, 14671-14676, 1992.

Saunders, C. P. R., Reply, J. Geophys. Res., 98, 10,823-10,825, 1993.

Saunders, C. P. R., Reply, J. Geophys. Res., 100, 1507-1510, 1995.

Seimon, A., Anomalous cloud-to-ground lightning in an F5 tornado-producing supercell thunderstorm on 28 August 1990, Bull. Amer. Meteor. Soc., 74, 189-203, 1993.

Simpson, G. C., and F. J. Scrase, The distribution of electricity in thunderclouds, Proc. R. Soc. London, Ser. A, 161, 309-352, 1937.

Stolzenburg, M., Characteristics of the bipolar pattern of lightning locations observed in 1988 thunderstorms, Bull. Amer. Meteor. Soc., 71, 1331-1338, 1990.

Stolzenburg, M., Observations of high ground flash densities of positive lightning in summertime thunderstorms, Mon. Wea. Rev., 122, 1740-1750, 1994.

Takahashi, T., Riming electrification as a charge generation mechanism in thunderstorms, J. Atmos. Sci., 35, 1536-1548, 1978.

Uman, M. A., The Lightning Discharge, 377 pp., Academic Press, Orlando, FL, 1987.

Vivekanandan, J., V. N. Bringi, M. Hagen, and P. Meischner, Polarimetric radar studies of atmospheric ice particles, IEEE Trans. on Geoscience and Remote Sensing, 32, 1-10, 1994.

Weinheimer, A. J., and A. A. Few, The electric field alignment of ice particles in thunderstorms, J. Geophys. Res, 92, 14,833-14,844, 1987.

Weisman, M. L., and J. B. Klemp, The dependence of numerically simulated convective storms on vertical wind shear and buoyancy, Mon. Wea. Rev., 110, 504-520, 1982.

Weisman, M. L., and J. B. Klemp, The structure and classification of numerically simulated convective storms in directionally varying wind shears, Mon. Wea. Rev., 112, 2479-2498, 1984.

Williams, E. R., The tripole structure of thunderstorms, J. Geophys. Res., 94, 13,151- 13,167, 1989.

Williams, E. R., M. E. Weber and R. E. Orville, The relationship between lightning type and convective state of thunderclouds. J. Geophys. Res., 94, 13,213-13,220, 1989a.

Williams, E. R., M. E. Weber and C. D. Engholm, Electrical characteristics of microburst- producing storms in Denver, Proc. 24th Conf. Radar Meteorology, Tallahassee, FL, pp. 89-92, Amer. Meteor. Soc., Boston, Mass., 1989b.

Williams, E. R., R. Zhang, and J. Rydock, Mixed phase microphysics and cloud electrification, J. Atmos. Sci., 48, 2195-2203, 1991.

Williams, E. R., and R. Zhang, Comment on "The effect of liquid water on thunderstorm charging" by C. P. R. Saunders et al., J. Geophys. Res., 98, 10,819-10,821, 1993.

Williams, E. R., Comment on "Thunderstorm electrification laboratory experiments and charging mechanisms," by C. P. R. Saunders, J. Geophys. Res., 100, 1503-1505, 1995.

Wilson, C. T. R., Investigations on lightning discharges and on the electric field of thunderstorms, Philos. Trans. A, 221, 73-115, 1920.

Zrnic*, D. S., V. N. Bringi, N. Balakrishnan, K. Aydin, V. Chandrasekar and J. Hubbert, Polarimetric measurements in a severe hailstorm. Mon. Wea. Rev., 121, 2223-2238, 1993.

Top of Page

Web Page Design by Ms. Brenda Dolan, Colorado State University