Abstract

The first flash produced by a storm usually does not strike ground, but little has been published concerning the time after the first flash before a cloud-to-ground flash occurs, particularly for a variety of climatological regions. To begin addressing this issue, this study analyzed data from very-high-frequency (VHF) lightning mapping systems, which detect flashes of all types, and from the U.S. National Lightning Detection Network (NLDN), which identifies flash type and detects roughly 90% of cloud-to-ground flashes overall. VHF mapping data were analyzed from three regions: north Texas, Oklahoma, and the high plains of Colorado, Kansas, and Nebraska. The percentage of storms in which a cloud-to-ground flash was detected in the first minute of lightning activity varied from 0% in the high plains to 10%–20% in Oklahoma and north Texas. The distribution of delays to the first cloud-to-ground flash varied similarly. In Oklahoma and north Texas, 50% of storms produced a cloud-to-ground flash within 5–10 min, and roughly 10% failed to produce a cloud-to-ground flash within 1 h. In the high plains, however, it required 30 min for 50% of storms to have produced a cloud-to-ground flash, and 20% produced no ground flash within 1 h. The authors suggest that the reason high plains storms take longer to produce cloud-to-ground lightning is because the formation of the lower charge needed to produce most cloud-to-ground flashes is inhibited either by delaying the formation of precipitation in the mid- and lower levels of storms or by many of the storms having an inverted-polarity electrical structure.

1. Introduction

The first flash produced by a storm usually does not strike ground (i.e., it is a cloud flash), but little information has been published concerning the distribution of times to the first cloud-to-ground flash (also called a ground flash) following the onset of lightning activity (MacGorman and Rust 1998, 229–230). Such information is important, because it would provide clues to the special conditions necessary for producing cloud-to-ground (CG) lightning, beyond the overall electrification necessary for producing any kind of lightning. It also affects how rapidly thunderstorms can be detected for hazardous weather operations by systems that mainly detect ground flashes.

One reason for the lack of such information has been the technical difficulty involved in detecting lightning flashes well in enough storms to generate reliable statistics. Prior to 1998, systems for mapping all types of lightning quickly and reliably covered only one relatively small region around the Kennedy Space Center permanently and covered other regions only for special observational periods. The technology for such systems has matured considerably in recent years and is now available routinely in a few regions, as well as during field programs elsewhere. Cloud-to-ground flash detection has been provided for the contiguous United States by the National Lightning Detection Network (NLDN) since around 1990 (Orville 2008; Cummins and Murphy 2009). In 2005, this network was modified to also detect 10%–20% of cloud flashes over Oklahoma, and this modification was extended nationally to provide a typical cloud flash detection efficiency of 10% the following year (Cummins and Murphy 2009). Although 10% is not a large enough percentage of cloud flashes to determine the distribution of time lags from first flash to the first cloud-to-ground flash, it will affect how rapidly the NLDN detects storms.

This study examines the distribution of times to the first cloud-to-ground flash by analyzing the timing of flashes detected by the NLDN relative to the timing of flashes detected by three-dimensional lightning mapping systems that map essentially all flashes that have traditionally been detectable by standard means, such as surface electric field change meters. The data presented in this paper were acquired May–August 2005 in Oklahoma and north Texas and May–July 2000 in Colorado, Kansas, and Nebraska.

2. Instrumentation

Total lightning data (cloud flashes plus ground flashes) were provided by three very-high-frequency (VHF) mapping systems (Fig. 1). 1) In central Oklahoma, data were provided by the Oklahoma Lightning Mapping Array (OK-LMA) built by New Mexico Institute of Mining and Technology (NMIMT) and operated jointly by the University of Oklahoma and the National Severe Storms Laboratory (MacGorman et al. 2008). 2) Total lightning data for the high plains of northwest Kansas, northeast Colorado, and southwest Nebraska were provided by a Lightning Mapping Array (LMA) operated by NMIMT during the Severe Thunderstorm Electrification and Precipitation Study (STEPS) field program in May–July 2000 (Rison et al. 1999; Thomas et al. 2004; Lang et al. 2004). (The operational characteristics of this system were essentially the same as those of the OK-LMA.) 3) Total lightning data in north Texas were provided by the Lightning Detection and Ranging II system (LDARII) operated by Vaisala around Dallas–Fort Worth (DFW; Carey et al. 2005). Each VHF mapping system contained a network of stations that measured the time and signal amplitude of the largest signal radiated by lightning in an unused VHF television band during each 80–100-μs period. From the differences in time of arrival between stations, the system computes the time and the three-dimensional location at which a lightning channel segment radiated the signal (Rison et al. 1999; Thomas et al. 2004).

Fig. 1.

Map of the three VHF total lightning mapping arrays used to determine the time of the first flash a storm produced. A circle indicates the approximate area of coverage by each network. The Dallas–Fort Worth network had fewer stations and a smaller area of coverage. The storms analyzed for this study were within each region of coverage.

Fig. 1.

Map of the three VHF total lightning mapping arrays used to determine the time of the first flash a storm produced. A circle indicates the approximate area of coverage by each network. The Dallas–Fort Worth network had fewer stations and a smaller area of coverage. The storms analyzed for this study were within each region of coverage.

These systems typically map tens to thousands of channel segments per flash to reveal the extent, structure, and development of the flash. Within the perimeter of each network (the inner 15%–25% of the area of each circle in Fig. 1), essentially all lightning is mapped in three dimensions with an accuracy of ≲50 m (Thomas et al. 2004). Location accuracy, particularly in the vertical dimension, degrades with range outside the perimeter of a network, but the horizontal error was no more than 2 km within the 200-km maximum range of our analysis and was adequate for this study.

All lightning ground strike data analyzed in this study are from the NLDN alone, unaided by VHF mapping systems. Cummins et al. (1998) describe the general characteristics of the NLDN, as it existed from 1995 to 2002. Much about the system remains the same, but in 2002–03, the station spacing was made somewhat denser and station sensitivity was improved (Cummins and Murphy 2009). In 2004, cloud flash detection was added in Oklahoma to the original capability for detecting cloud-to-ground flashes, and in April 2006, this capability was extended nationwide (Biagi et al. 2007; Cummins and Murphy 2009). Regional observations and theoretical extensions of observational results indicate that the NLDN detects in excess of 90% of cloud-to-ground lightning flashes overall in the contiguous United States (Biagi et al. 2007; Fleenor et al. 2009; Cummins and Murphy 2009). Because cloud flashes tend to radiate weaker signals than ground flashes do in the frequency band at which the NLDN operates, the detection efficiency for cloud flashes depends more sensitively on station spacing. For the NLDN, typical values of detection efficiency for cloud flashes are estimated to be 10% over the contiguous United States, significantly less than those for ground flashes. For the configuration of NLDN stations used in 2005 for this study, typical cloud flash detection efficiency is estimated to have been 10%–20%.

3. Analysis procedure

The procedure for this study was to peruse data from the VHF mapping systems alone to identify periods of lightning activity. Each storm during these periods was then examined in detail to identify when a VHF system first detected a flash and when the NLDN detected its first flash in various categories, which will be described shortly. To try to ensure that the VHF mapping systems detected the first flash, the range of storms for this study was constrained to be completely within roughly 200 km of the center of the LMA and OK-LMA networks and within roughly 150 km of the LDARII for at least the first hour of lightning activity (the DFW network was smaller and had somewhat less sensitivity than the other two networks). The lightning activity of each storm was tracked manually to ensure that, once a storm had produced lightning, its later lightning would not be considered a first flash of another storm. To be considered a new storm, the storm’s lightning had to be ≳25 km away from all recent lightning activity and clearly separate from a rough manually extrapolated position of previous lightning activity.

For the analysis of lightning in Oklahoma, three categories of NLDN flashes were used to examine the effect on storm detection of including NLDN detection of cloud flashes. Because of changes made by 2004 to increase the NLDN system’s sensitivity, it was found that many flashes identified in the southern plains and high plains as ground flashes having peak currents (Ipk) <10 kA actually were cloud flashes, regardless of the polarity of charge lowered to ground (Johnson and Mansell 2006; Fleenor et al. 2009). In the present NLDN, positive ground flashes (+CG flashes) with small peak currents are categorized as cloud flashes, but negative ground flashes (−CG flashes) with small peak currents are kept as ground flashes. To try to consider these issues, our analysis treated −CG flashes detected by the NLDN with peak currents <10 kA as a separate category from ground flashes having larger peak currents and included +CG flashes detected by the NLDN with peak currents <10 kA in the cloud flash category. (We selected the 10-kA threshold based on previous analyses, but later found it is smaller than the 15-kA threshold for +CG classification that the NLDN now uses operationally.) This study did not tabulate ground flash polarity for NLDN flashes with peak currents ≥10 kA, but categorized all of them simply as CG flashes.

Our analysis determined the time lag from the first flash detected by the VHF mapping system to the first flash of various categories of flashes detected by the NLDN. The first category of NLDN flashes consisted of all ground flashes with peak currents ≥10 kA. The second category consisted of the first category plus all negative ground flashes with peak currents of <10 kA. The third consisted of the first two plus all NLDN-detected cloud flashes. Because the NLDN’s cloud flash detection was available for this study only within the coverage of the OK-LMA, cloud flash detection was not analyzed for north Texas or for the high plains. Note that, because the categories were cumulative, the time lags for higher categories were always less than or equal to the time lags for lower categories.

Because the initial flash rate obviously influences the time lag, we also tabulated the total flash rate within 5-min of the first flash. In one analysis, the distributions were sorted into 3 groups based on the average total flash rate for the first 5 min of lightning activity in each storm: “low” for storms producing <1 flash per minute, “medium” for storms producing ≥1 and <3 flashes per minute, and “high” for storms producing ≥3 flashes per minute. Results from the OK-LMA also were sorted by the month in which the storm occurred to try to examine systematic variations across the analyzed part of the warm season, May–August 2005.

Besides looking at the time lag from the first flash detected by a VHF system to the first flash in each NLDN category, we were also interested in how fast lightning activity increased, and so analyzed the lag until the first period in which the storms produced ≥2 flashes per 5 minutes and ≥5 flashes per 5 minutes in the VHF data and in the various categories of NLDN flashes. Some storms never reached these thresholds of flash rate.

Our analysis of storms in north Texas and in the high plains differed somewhat from our analysis of storms in Oklahoma. For one thing, NDLN cloud flash data were unavailable in those two regions during the analyzed period. Because of this and because Cummins et al. (1998) had found that most of the NLDN-detected +CG flashes with peak currents <10 kA were actually from cloud flashes, our analysis ignored +CG flashes with small peak currents in those two regions. As for our analysis of Oklahoma data, the only +CG flashes included in the ground flash category were those having peak currents ≥10 kA. However, all −CG flashes in those 2 regions were included, regardless of peak current, in part because Fleenor et al. (2009) found that cloud flashes were mistaken as −CG flashes much less often in north Texas than in Oklahoma and in part because the high plains dataset preceded the 2004 modification of the NLDN.

4. The additional time required for storms to produce various types of lightning

a. Cloud flash and cloud-to-ground flash detection in Oklahoma

Figure 2 shows the elapsed time to the first flash in each of the various categories of NLDN flashes (all CG flashes with peak current ≥10 kA, adding −CG flashes with peak current <10 kA, and adding NDLN detected cloud flashes) following the first flash in storms 1 May–15 August 2005 in Oklahoma. Note that 20% of storms produced a ground flash with a peak current of at least 10 kA within the first minute of lightning activity (24% if flashes with small negative peak currents were included). It required 6 min for 50% of storms to produce a ground flash and 15 min for 75% to do so (4 and 12 min, respectively, with small negative peak currents included). By the end of 1 h, 88% of storms had produced a ground flash with a peak current of at least 10 kA (93% when small negative peak currents were included).

Fig. 2.

Cumulative probability of the NLDN detecting a flash in each category following the first flash detected by the OK-LMA in Oklahoma May–August 2005. All flash rates in this paper were determined in 1-min increments of elapsed time.

Fig. 2.

Cumulative probability of the NLDN detecting a flash in each category following the first flash detected by the OK-LMA in Oklahoma May–August 2005. All flash rates in this paper were determined in 1-min increments of elapsed time.

An additional 5%–10% of storms produced a flash detected by the NLDN at each elapsed time when cloud flashes were included, and so obviously shortened the time required to reach any percentile of storms, especially at the larger percentiles. With NLDN cloud flashes added, 50% of storms produced a flash detected by the NLDN within 3 min, 75% did so within 8 min, and 90% did so within 15 min. By the end of 1 h, 97% of storms had produced a flash detected by the NLDN.

b. Variations relative to initial flash rate and to warm-season month

One issue was to determine how stable each distribution was from month to month in 2005 in Oklahoma. Figure 3 shows the month-to-month variation in the fraction of storms producing their first flash in each category within any given elapsed time. Note that the fraction of storms that was initially detected, the fraction eventually detected, and the rate of increase in the fraction all varied considerably from month to month. One likely reason for this variability is typical systematic differences in the flash rates produced by storms from month to month, as will be discussed later.

Fig. 3.

Cumulative probability of the NLDN detecting a flash in each category following the first flash detected by the OK-LMA for each month during the period studied in 2005.

Fig. 3.

Cumulative probability of the NLDN detecting a flash in each category following the first flash detected by the OK-LMA for each month during the period studied in 2005.

To examine the influence of flash rate on the time to the first flash in each category, the initial 5-min flash rates of the storms were sorted into low, medium, or high flash rates, as described in the previous section, and the cumulative probabilities were recomputed as a function of elapsed time for each range of initial flash rates (Fig. 4). Note that the distributions changed systematically as the initial flash rate progressed from low to high. Both the maximum fraction of storms detected and the rapidity with which the larger fractions were reached tended to increase as one progressed from low to high.

Fig. 4.

Cumulative probability of the NLDN detecting a flash in each category following the first flash detected by the OK-LMA, shown separately for three different ranges of initial total flash rate (5-min average rate).

Fig. 4.

Cumulative probability of the NLDN detecting a flash in each category following the first flash detected by the OK-LMA, shown separately for three different ranges of initial total flash rate (5-min average rate).

However, there appears to be one anomaly. During the first few minutes of the high flash rate storms, the fraction of storms producing a ground flash actually was smaller than for storms with smaller initial flash rates. The reason is likely related to the observation by MacGorman et al. (1989) that ground flash activity was delayed in a severe storm that had very strong updrafts. Several studies have shown that total flash rates tend to increase with increasing updraft mass flux through the mixed phase region (e.g., Wiens et al. 2005; Kuhlman et al. 2006) and that ground flash rates tend to increase as precipitation forms at the middle levels of storms and descends (e.g., Goodman et al. 1988; MacGorman et al. 1989). Larger updraft speeds can increase the time required for precipitation to form at midlevels and descend to the lower altitudes at which charge is thought to be needed to initiate most ground flashes (e.g., Jacobson and Krider 1976; MacGorman et al. 1989, 2001; Mansell et al. 2002, 2005).

As one might expect, NLDN-detected cloud flashes made the biggest difference for low-flash-rate storms, for which an additional 5%–15% of storms were added at each elapsed time. Small-peak-current negative ground flashes also contributed 5%–13% of storm detections for low-flash-rate storms. Recent analyses of OK-LMA data have shown that the upgraded NLDN identifies some cloud flashes as small-peak-current negative ground flashes in Oklahoma storms, probably a result of changes made to increase the NLDN’s sensitivity. These studies have suggested that misidentification happens most often in severe storms and in storms whose ground flash activity is dominated by storms that lower positive charge to ground, instead of the usual negative charge (Johnson and Mansell 2006; Biagi et al. 2007). The actual contribution of cloud flashes to storm detection, if the NLDN correctly identified the category of all flashes, probably would be somewhere between the contribution of NLDN detections identified as cloud flashes and the contribution if small-peak-current negative ground flashes were added to the cloud flash category.

c. Regional variations in the time to first ground strike

As mentioned previously, another concern is the effect of varying climatologies on results. To begin analyzing this issue, we analyzed data from two other regions described in the instrumentation section: 1) data from north Texas and 2) data collected previously from the high plains during the STEPS field program.

In north Texas (Fig. 5), the fraction of storms producing a ground flash during the first 30 min of lightning activity was smaller than in Oklahoma. Only 12% of storms produced a ground flash detected by the NLDN within 1 min of the first flash detected by the LDARII. Furthermore, it took 8 min for 50% of storms to produce a ground flash and 23 min for 75% to do so, both times substantially longer than required in Oklahoma. However, the 89% of storms that produced a ground flash within 1 h in north Texas was comparable to the 88%–93% that did so in Oklahoma.

Fig. 5.

Cumulative probability of the NLDN detecting its first cloud-to-ground flash following the first flash detected by the LDARII in north Texas. Positive ground flashes with peak currents less than 10 kA were omitted from the NLDN dataset, because they are now categorized as cloud flashes. NLDN cloud flash data were not available for these storms.

Fig. 5.

Cumulative probability of the NLDN detecting its first cloud-to-ground flash following the first flash detected by the LDARII in north Texas. Positive ground flashes with peak currents less than 10 kA were omitted from the NLDN dataset, because they are now categorized as cloud flashes. NLDN cloud flash data were not available for these storms.

On the high plains (Fig. 6), it took much longer for a storm to produce its first cloud-to-ground flash after its first cloud flash. The NLDN detected no ground flash within the first minute of lightning activity. It required 31 min for 50% of storms to produce a ground flash and 44 min for 75% to do so. Only 80% of storms produced a ground flash detected by the NLDN within 1 h. Previous studies have found that cloud-to-ground flashes tend to compose a smaller fraction of lightning in storms on the high plains (e.g., Boccippio et al. 2001), so one would expect it to take longer for these storms to produce a ground flash, and a smaller fraction of the storms would be expected to produce a ground flash within 1 h, as found by our analysis.

Fig. 6.

Cumulative probability of the NLDN detecting its first cloud-to-ground flash following the first flash of any type detected by the lightning mapping array (Thomas et al. 2004) used in the STEPS field program in 2000. NLDN cloud flash data were not available for these storms.

Fig. 6.

Cumulative probability of the NLDN detecting its first cloud-to-ground flash following the first flash of any type detected by the lightning mapping array (Thomas et al. 2004) used in the STEPS field program in 2000. NLDN cloud flash data were not available for these storms.

To make it easier to compare results among these three regions, a summary of the above results is shown in Table 1. Keep in mind that these statistics are all relative to the first flash detected by the LMA or LDARII total-lightning mapping system, which was assumed, therefore, to have detected the first flash in all storms at an elapsed time of 0 min.

Table 1.

The elapsed time required for storms to produce the first ground flash or other NLDN-detected flash relative to the time of the first flash detected by a VHF total lightning mapping system for storms in Oklahoma, north Texas, and the high plains.

The elapsed time required for storms to produce the first ground flash or other NLDN-detected flash relative to the time of the first flash detected by a VHF total lightning mapping system for storms in Oklahoma, north Texas, and the high plains.
The elapsed time required for storms to produce the first ground flash or other NLDN-detected flash relative to the time of the first flash detected by a VHF total lightning mapping system for storms in Oklahoma, north Texas, and the high plains.

d. Elapsed time for Oklahoma storms to produce various flash rates

We also examined how rapidly various types of flash rates increased in Oklahoma storms. Besides this topic being of interest to better understand lightning production, studies have found that trends in total flash rates are related to trends in precipitating ice volume and mass and to increasing updraft volume and mass flux through the mixed phase region, as well as to increasing potential for hazardous weather (e.g., Lhermitte and Krehbiel 1979; MacGorman et al. 1989; Williams et al. 1999; Wiens et al. 2005; Schultz et al. 2009). To examine fully how well various subsets of lightning replicate actual total flash trends would require a very laborious analysis of several tens of minutes of data for each storm, an analysis beyond the limits of time and labor available for the overall study. However, one condition necessary for a lightning dataset to produce trends in a timely manner is to detect frequent enough flashes to depict the trends.

To begin addressing the relative timing of trends for various types of lightning, we analyzed how long it took for storms in the analyzed region to produce two thresholds of flash rates for flashes detected by the OK-LMA and for each category of flash detected by the NLDN: ≥2 flashes per 5 minutes and ≥5 flashes per 5 minutes. The algorithm used to delineate flashes from OK-LMA data is described by MacGorman et al. (2008) and Lund et al. (2009). The flash rates determined by the algorithm from OK-LMA data are probably close to the actual flash rates when flash rates are less than 60 min−1 (MacGorman et al. 2008), as they were for all periods we examined for this study.

Results for the threshold of 2 flashes per 5 minutes are shown in Fig. 7 and Table 2. The total flash rates detected by the OK-LMA reached this threshold within 1 min in only a little more than 40% of storms. (This meant that storms produced a second flash within the first minute.) It took only 5 min, however, for this rate to be produced in 90% of storms and 19 min to be produced in 99% of storms. Flashes detected by the NLDN are a subset of those detected by the OK-LMA, so the flash rates detected by the NLDN were usually smaller, as expected. If only NDLN-detected ground flashes with peak currents ≥10 kA were included, very few storms (≲l%) produced ≥2 flashes per 5 minutes in the first minute of lightning activity. It took 14 min for 50% of storms to produce 2 ground flashes within 5 min and took a little less time (11 min) to do so if negative ground flash detections with peak currents <10 kA were considered to be actual ground flashes.

Fig. 7.

Cumulative probability of the time required after the first flash for storms to produce at least 2 flashes per 5 minutes in each category of flashes detected by the NDLN and the OK-LMA in 2005.

Fig. 7.

Cumulative probability of the time required after the first flash for storms to produce at least 2 flashes per 5 minutes in each category of flashes detected by the NDLN and the OK-LMA in 2005.

Table 2.

The time required for storms to produce 2 or 5 flashes per 5 minutes in Oklahoma for the total flashes detected by the OK-LMA and the various NLDN flash categories. The values in parentheses for each parameter give the monthly values (May, June, July, and August) corresponding to the single value listed above the parentheses for the entire analysis period.

The time required for storms to produce 2 or 5 flashes per 5 minutes in Oklahoma for the total flashes detected by the OK-LMA and the various NLDN flash categories. The values in parentheses for each parameter give the monthly values (May, June, July, and August) corresponding to the single value listed above the parentheses for the entire analysis period.
The time required for storms to produce 2 or 5 flashes per 5 minutes in Oklahoma for the total flashes detected by the OK-LMA and the various NLDN flash categories. The values in parentheses for each parameter give the monthly values (May, June, July, and August) corresponding to the single value listed above the parentheses for the entire analysis period.

Note that some storms did not attain these rates even when considering all lightning, and this placed an upper bound on the fraction attaining these rates with NLDN-detected flashes. To examine the behavior of NLDN-detected flashes relative to total lightning activity, we divided the number of storms that satisfied the threshold flash rate within each time interval in NLDN data by the number of storms that satisfied the threshold in the same time interval in OK-LMA data (not shown, but equal to the LMA curve divided by each of the NLDN curves in Fig. 7). It took 13 min for NLDN ground flashes alone to attain a flash rate of ≥2 flashes per 5 minutes in 50% of the storms having this total flash rate in OK-LMA data, and took only 7 min if NLDN-detected cloud flashes were included. Furthermore, 22% of the storms that produced this flash rate in OK-LMA data within 1 h failed to produce this flash rate in ground flashes having peak currents ≥10 kA in NLDN data, and 12% failed to do so when all NLDN-detected flashes were included.

Again, we examined the variability in these results by examining each month separately (Fig. 8 and Table 2). In OK-LMA data, the cumulative fraction of storms producing 2 flashes of any sort within a 5-min period behaved similarly in all 4 months, the most pronounced difference being in the first minute of lightning activity in May and June. However, in NLDN ground flash data, the cumulative fraction varied considerably from month to month. At almost every elapsed time, the fraction of storms producing ≥2 ground flashes per 5 minute period in May was substantially smaller than the fraction in other months. In the other 3 months, the fraction was fairly similar during the first 10 min of elapsed time, but then the fraction increased more rapidly at later times in June and August, eventually reaching roughly 90%, while not quite reaching 80% in July. The progression of distributions in May, July, and August might lead one to suspect a systematically increasing trend by season, but the fraction of storms in June is greater than the fraction in July for all times after 20 min of elapsed time and approaches the values observed in August. It may be that much of the departure from monotonically increasing distributions from May to August was caused by interannual variations in persistent synoptic patterns within one or more months.

Fig. 8.

Cumulative probability of the time required after the first flash for storms to produce at least 2 flashes per 5 minutes in each category of flashes detected by the NLDN and the OK-LMA during each month of this study in 2005.

Fig. 8.

Cumulative probability of the time required after the first flash for storms to produce at least 2 flashes per 5 minutes in each category of flashes detected by the NLDN and the OK-LMA during each month of this study in 2005.

Figure 9 and Table 2 show results when the threshold was 5 flashes per 5 minutes. As one might expect, it typically required more time for storms to attain this threshold than to attain the lesser threshold. In total lightning data, at least 5 flashes per 5 minutes were produced in 50% of storms within 5 min, in 75% of storms within 26 min, and in only 80% of storms within 1 h. Ground flashes did not attain this rate at all within the first 2 min of lightning activity. When using only ground flashes having peak currents ≥10 kA, the flash rate threshold was reached in 25% of storms within 19 min, and the 50th percentile was not attained within 1 h, the maximum percentage being 47%. Adding negative ground flashes having small peak currents typically increased the percentage of storms attaining this threshold by ≤5% points at most elapsed times, with approximately 50% of storms attaining it within 1 h. NLDN cloud flash data typically contributed only an additional 2%–5% points to the fraction of storms detecting 5 flashes in 5 minutes: when these data were included, the flash rate threshold was reached in 25% of storms within 15 min, in 50% of storms within 46 min, and in 53% of storms within 1 h.

Fig. 9.

Cumulative probability of detecting 5 flashes within 5 minutes when using each lightning dataset, as a function of the elapsed time following the first flash detected by the OK-LMA.

Fig. 9.

Cumulative probability of detecting 5 flashes within 5 minutes when using each lightning dataset, as a function of the elapsed time following the first flash detected by the OK-LMA.

The percentages in the previous paragraph were relative to all analyzed storms. As noted previously, NLDN-detected flashes are a subset of all flashes, so the percentages for NLDN-detected flashes are larger when compared only with storms satisfying the flash rate threshold in OK-LMA data (not shown, but equal to the LMA curve divided by each of the NLDN curves in Fig. 9): 50% of storms having ≥5 flashes per minute for all types of lightning also satisfied this threshold flash rate for NLDN-detected ground flashes with peak currents ≥10 kA within 31 min, and 59% satisfied the threshold ground flash rate within 1 h. If we consider all flashes, including cloud flashes, detected by the NLDN, 50% of storms having this flash rate in OK-LMA data also had this flash rate in NLDN data within 20 min, and 66% did so within 1 h.

Figure 10 and Table 2 show how lightning behavior in attaining a threshold of 5 flashes per 5 minutes varied from month to month in 2005. As for the lesser threshold, the fraction of storms producing this flash rate in May was much smaller than in the other analyzed months: the fraction attaining this threshold of total flash rates in May was approximately two-thirds the fraction in other months, and the fraction attaining this rate using only ground flashes was even smaller. The distributions in July and August were similar, but the distribution in June was somewhat different. In June, the distribution for total lightning flashes was somewhat larger than in other months after the first 3 min of elapsed time, and the distribution for ground flashes increased eventually to larger values than in other months, reaching roughly 70% within 1 h in June versus 45%–55% in July and August. Without a dataset extending over several years, we cannot determine how much of these monthly variations is due to typical seasonal trends and how much is influenced by interannual variations about the mean for a given month.

Fig. 10.

Cumulative probability of the time required after the first flash for storms to produce at least 5 flashes per 5 minutes when using each lightning dataset (the 3 categories of flashes detected by the NLDN and flashes detected by the OK-LMA) during each month of this study in 2005.

Fig. 10.

Cumulative probability of the time required after the first flash for storms to produce at least 5 flashes per 5 minutes when using each lightning dataset (the 3 categories of flashes detected by the NLDN and flashes detected by the OK-LMA) during each month of this study in 2005.

5. Discussion and conclusions

a. Observations of the timing of a storm’s first cloud-to-ground flash

This study examines cloud-to-ground lightning production relative to total lightning production. Most previous observations had found that production of cloud-to-ground lightning almost always followed the production of cloud flashes by several minutes (e.g., MacGorman and Rust 1998, 229–230), with rare exceptions (Krehbiel 1986). However, the advent of systems capable of mapping lightning over hundreds of kilometers has made it possible to document this tendency more carefully. We used VHF lightning mapping systems in Oklahoma, north Texas, and the high plains to indicate the timing of total lightning activity, regardless of flash type, and used data from the NLDN to indicate the timing of cloud-to-ground flashes.

In May–August 2005 in Oklahoma, 20%–24% of storms produced a NLDN-detected ground flash within the first minute of lightning activity, 50% of storms did so within 4–6 min, 75%, within 12–15 min, and 88%–93%, within 1 h. (The actual value for ground flashes depends on how many of the small-peak-current −CG flashes indicated by the NLDN were valid ground flashes, not misidentified cloud flashes, a number we could not estimate within the scope of our project.) In north Texas during the same period, 12% of storms produced a ground flash in the first minute; 50% of storms did so within 8 min; 75%, within 23 min; and 89%, within 1 h.

The time interval from the beginning of lightning activity to the first cloud-to-ground flash tended to be considerably longer for storms on the high plains than for storms in the other two regions. None of the high plains storms we analyzed produced a ground flash in the first 2 min of lightning activity, and only 80% did so within 1 h. It required 31 min after lightning began for 50% of storms to produce a ground flash, and 44 min for 75% to do so. This is consistent with the climatological observation by Boccippio et al. (2001) that ground flashes compose a smaller fraction of lightning flashes in the high plains than in most other regions of the contiguous United States.

We used the extensive 2005 dataset from Oklahoma to examine variations in the relative timing of the first cloud-to-ground flash. The overall tendency after the first few minutes of lightning activity was for the fraction of storms producing ground flashes within any particular elapsed time to increase each month from May to August. However, the fraction of storms producing ground flashes during the first few minutes was more erratic, being smallest in May and July and largest in August.

Some of the variability was due to a systematic dependence on the initial total flash rates of storms, averaged over the first 5 min of lightning activity. The higher percentiles of storms producing ground flashes tended to be reached more quickly as the initial total flash rate increased, as one might expect. One exception to this tendency occurred during the first two minutes of lightning activity for the category with the highest initial total flash rates, possibly because some severe storms tend to delay ground flashes, as noted, for example, by MacGorman et al. (1989). They suggested the delay was caused by the strong updrafts of severe storms elevating the lower precipitation higher than in most storms, thereby either elevating or delaying the formation of the lower charge needed to produce ground flashes. Some of the variability from month to month and during the first two minutes of lightning activity probably was due also to systematic seasonal differences in storm characteristics (such as the amount and relative speed at which precipitation appeared near the freezing level) affecting the timing of initial cloud-to-ground flashes, but analyzing the nature of this dependence was beyond the scope of this project.

Because NLDN data are often used to detect thunderstorms, we used the data from Oklahoma to examine how much the timeliness of storm detection was improved by including the limited NLDN detection of cloud flashes. (Oklahoma was the only location having both NLDN cloud flash data and VHF lightning mapping data during data collection for this study, although cloud flash data are available nationally now.) Including the NLDN’s cloud flash data allowed an additional 5%–10% of storms to be detected at each elapsed time and decreased the time required to detect a given percentage of storms: 50% of storms were detected within 3 min (vs 4–6 min without cloud flash data); 75%, within 8 min (vs 12–15 min); and 90%, within 15 min (vs at least 29 min). However, the contribution of NLDN cloud flash data to thunderstorm detection varied considerably from month to month. The contribution of cloud flashes was greatest in May, when the contribution by ground flashes was smallest, and was smallest in August, when the contribution by ground flashes was largest. There was a similar dependence of the relative contribution of NLDN ground flashes and NLDN cloud flashes on the average total flash rate during the first 5 min of storms; the contribution of NLDN cloud flashes to storm detection was greatest for storms having the smallest initial total flash rate and tended to be smallest for storms having the largest initial flash rate.

b. Observations of the timing of larger flash rates

We also analyzed how rapidly various categories of flash rates increased in storms. As noted previously, trends in total flash rates appear to be correlated with trends in precipitating ice volume and mass, with trends in updraft volume and mass flux through the mixed phase region, and with changes in storm severity (e.g., Lhermitte and Krehbiel 1979; MacGorman et al. 1989; Williams et al. 1999; Wiens et al. 2005; Schultz et al. 2009). [In contrast, studies such as Reap and MacGorman (1989) have found that peaks in −CG flash rates in the southern plains have little relationship to storm severity and updraft intensity. MacGorman et al. (1989), Carey and Rutledge (1996), and MacGorman et al. (2007) suggested rather that trends in ground flash rates are associated with the formation and descent of precipitation to lower levels of storms, as noted above.] Such studies suggest that forecasters may be able to infer storm characteristics such as updraft trends from trends in some types of lightning flashes.

To examine fully how well various less complete datasets replicate actual total flash rate trends would require a very laborious analysis of several tens of minutes of data for each storm. However, one condition necessary for a lightning dataset to produce trends in a timely manner is to detect frequent enough flashes to depict the trends. To begin examining how rapidly flash rates increase for various categories of flashes, we analyzed how long it took for storms to produce two thresholds of flash rates, ≥2 flashes per 5 minutes and ≥5 flashes per 5 minutes, for flashes detected by the OK-LMA and for each analyzed category of flash detected by the NLDN.

For the lightning detected by the OK-LMA, which included all types of flashes, storms produced ≥2 flashes per 5 minutes in a little more than 40% of storms within the first minute, in 50% of storms within 2 min, in 90% of storms within 5 min, and in 99% of storms within 20 min. Storms produced ≥5 flashes per 5 minutes in 5% of storms within a minute, in 50% of storms within 5 min, in 75% of storms within 26 min, and in 80% of storms within 1 h. The OK-LMA data probably depict close to the actual flash rates of the storms we analyzed and so provide an upper limit to the performance of mapping systems, such as the NLDN, that detect a smaller subset of flashes.

When using only NLDN ground flashes with estimated peak currents ≥10 kA, storms produced 2 flashes per 5 minutes in 1% of storms within 1 min, in 50% of storms within 14 min, and in 77% of storms within 1 h. If all NLDN detected flashes were considered, including cloud flashes, storms produced this flash rate in 3% of storms within 1 min, in 50% of storms within 11 min, and in 87% of storms within 1 h. Note that these percentages are relative to all analyzed storms. If we consider only those storms that produced at least 2 flashes per 5 minutes in OK-LMA data, the fraction of storms producing that rate of ground flashes and of all flashes detected by the NLDN are, of course, larger. It required 13 min for half of the storms with a total flash rate of ≥2 flashes per 5 minutes to reach this threshold when using only cloud-to-ground lightning having peak currents ≥10 kA and required only 7 min if all NLDN flashes were included. When including only cloud-to-ground lightning having peak currents ≥10 kA, 22% of storms failed to produce this flash rate within 1 h, and when including all NLDN flashes, 12% failed to do so.

The percentage of storms producing the larger flash rate of 5 flashes per 5 minutes in NLDN data tended to be at least 20% points less than the percentage producing this rate in OK-LMA data for all elapsed times and all categories of flashes we examined. For total flashes from the OK-LMA, only 5% of storms produced 5 flashes per 5 minutes within 1 min, 25% of storms did so within 3 min, 50% of storms did so within 5 min, 75% did so within 26 min, and 80% did so within 1 h. When using only ground flashes with peak currents ≥10 kA, no storms produced this flash rate in the first 2 min, 25% did so within 19 min, and 47% did so within 1 h. When using all NLDN detected flashes, no storms produced this flash rate in the first 2 min, 25% did so within 15 min, 50% did so within 46 min, and 53% did so within 1 h.

Again, the fraction of storms detected by the NLDN was larger when considered as a subset of storms that produced 5 flashes per 5 minutes in the total flash dataset from the OK-LMA, instead of being a subset of all storms that produced lightning. When considering only those NLDN-detected flashes with peak currents ≥10 kA in this reduced ensemble of storms, 50% of storms produced this flash rate in 31 min, and 59% did so within 1 h. When considering all NLDN detected flashes in this reduced ensemble of storms, 50% of storms produced this flash rate in 20 min, and 66% did so within 1 h. Thus, even with cloud flash data, the NLDN has only limited capability to detect trends toward larger total flash rates, as one would expect from its relatively small cloud-flash detection efficiency. However, it still is possible that the cloud flashes the NLDN detects are preferentially keyed to storm intensification or severe weather in a way that would be useful to forecasters. Unfortunately, there was not enough severe weather within the region and period analyzed to test this possibility.

c. Influences on the relative timing and number of ground flashes

We suggest that most of the variation in the timing (and perhaps amount) of cloud-to-ground lightning relative to total lightning activity is caused by variations in the timing and amount of lower charge formed below a midlevel charge region of opposite polarity. MacGorman et al. (2007) noted that to produce cloud-to-ground lightning, not only must a storm be electrified enough overall to produce lightning, but it must also have a configuration of charge that produces a channel to ground. As mentioned in previous sections, several investigators have suggested that the configuration of charge that initiates a ground flash typically consists of a midlevel charge above a lower charge of opposite polarity. This configuration increases the electric field below the midlevel charge in a direction causing lightning propagation toward ground. Because the lower charge region often contains less charge than the midlevel charge region and so produces a shallower potential well, the downward propagation of lightning may not stop in the lower charge, but may continue to the ground (MacGorman et al. 1989, 2001; MacGorman and Rust 1998; Coleman et al. 2008).

In a storm having a normal-polarity electrical structure (i.e., midlevel negative charge above a smaller low-level positive charge), the present understanding is that having enough lower positive charge usually requires either 1) inductive charging of precipitation below the roughly −10°C isotherm after the midlevel negative charge region is well established or 2) noninductive charging of frozen precipitation below the roughly −10° to −15°C isotherm. This requirement is consistent with inferences by Larson and Stansbury (1974), Goodman et al. (1988), MacGorman et al. (1989, 2007), and Carey and Rutledge (1996) that production of ground flashes is typically associated with the formation and descent of precipitation to lower levels of storms, perhaps to somewhere between roughly 5° and −15°C.

Observations by Bruning et al. (2007) of a weak Oklahoma thunderstorm in which the first flashes were negative cloud-to-ground flashes bolster this understanding further. The only charges initially involved in lightning were positive charge at roughly 3°C and negative charge at −6.5° to −19°C. Bruning et al. (2007) and a modeling study by Mansell et al. (2010) both indicated that precipitation and charge density were initially weak higher in this storm, where a sizeable upper positive charge region normally is found and did form later. Thus, one scenario in which ground flashes are the earliest flashes in a storm appears to be in relatively weak storms initially having most charge lower in the storm, with much less charge in upper storm regions.

We expect, conversely, that the time from the initial flash to the first cloud-to-ground flash tends to be later on the high plains than in Oklahoma and north Texas because formation of the lower charge is delayed or inhibited systematically in storms on the high plains. Besides the evidence mentioned above, support for this expectation is provided, for example, by the complete lack of cloud-to-ground lightning in the high plains low-precipitation supercell storm on 3 June 2000, which produced substantial flash rates (up to almost 30 min−1; Rust and MacGorman 2002; Rust et al. 2005; Tessendorf et al. 2007a). As noted by Rust et al. and Tessendorf et al. none of the flashes observed in this storm involved a lower charge region, and Rust and MacGorman and Rust et al. found that, if a lower charge region existed, its charge density was much less than the charge density in the higher regions involved in lightning.

The delay or lessening of the electrification needed to produce the lower charge region may be the result of a combination of factors that vary from storm to storm:

  1. High cloud base and shallow warm-cloud layer. Storms with a high cloud base and a resulting shallow layer for warm rain processes will shift precipitation growth to colder altitudes than found in storms with a warmer cloud base (Williams et al. 2005; Carey and Buffalo 2007). As a result, graupel concentrations in the lower part of the updraft’s mixed-phase region, where graupel can gain a positive charge by noninductive charge exchange with cloud ice in a classical tripole charge distribution, will be less (in some cases nonexistent) relative to graupel concentrations in the updrafts of storms with a warmer cloud base. The lower positive charge carried by graupel would be similarly reduced.

  2. Unusually large concentrations of cloud condensation nuclei (CCN). Ziegler et al. (2010) show that a similar shift of precipitation growth to colder altitudes can be caused by extremely large CCN concentrations. Thus, large CCN concentrations would also greatly reduce the amount of charge in the lower positive charge region from what would usually be found in the majority of otherwise similar storms having typical CCN concentrations.

  3. Strong upper-level, storm-relative (UL-SR) winds. Based on their statistical study showing that supercell morphology tended to progress from low-precipitation (LP) to heavy-precipitation (HP) as UL-SR winds decreased, Rasmussen and Straka (1998) inferred that strong UL-SR winds (i.e., above 9 km MSL) tend to reduce precipitation growth. Because the time required for precipitation formation in a supersaturated parcel is thought to be 10–30 min (e.g., Young 1993), comparable to or greater than the residence time of hydrometeors in the strong updrafts of supercell storms, they suggested that most precipitation development in strong updrafts typically involves recirculating hydrometeors from older cells or regions. If UL-SR winds are large, however, most hydrometeors are advected too far downstream to be recirculated into the updraft, so precipitation development in the updraft is slowed greatly. The result would be the visually observed skeleton structure of LP supercell storms, with mostly small cloud particles visible in and surrounding the updraft through middle levels of the storm, topped by outflowing cloud and precipitation particles. Downshear from the updraft, large precipitation particles are detected on radar, but are too sparse to form the visible rain curtain seen in classic and HP supercell storms. Such a structure, with little precipitation in the lower mixed-phase region, would also inhibit or eliminate the formation of a lower positive charge.

  4. Pronounced bounded weak echo region (BWER). A BWER exists because the core updraft speed is so large that the altitude at which precipitation forms is higher than usual and precipitation then rises very quickly, detrains from the updraft, and descends around the updraft core, as observed for example by Payne et al. (2010). If the storm has a normal-polarity electrical structure, those graupel particles just above the BWER that are warm enough to gain positive charge during noninductive microphysical interactions will pass quickly into higher levels in which the graupel gains a negative charge (e.g., Takahashi and Miyawaki 2002; Saunders et al. 2006; Emersic and Saunders 2010), eventually reversing the polarity of graupel charge. If the reversal is fast enough, as would be likely in a very strong updraft, it will occur before sedimentation can separate positive graupel far enough from negative cloud ice to produce a sizeable lower positive charge region. If, on the other hand, enough lower positive charge should form above the bounded weak echo region to initiate downward-propagating lightning, the lightning still would be much farther than usual from the ground and would propagate through a region in which the electric potential is climbing out of a potential well, thereby lessening the electric field driving new propagation farther downward and lessening the probability of the lightning reaching the ground (MacGorman et al. 1989, 2001; MacGorman and Rust 1998; Mazur and Ruhnke 1998).

  5. Updrafts with an inverted-polarity electrical structure (i.e., a structure having an upper-level negative charge above a midlevel positive charge). Several who have observed inverted-polarity storms have suggested that the polarity was inverted because updrafts realized unusually large liquid water contents (e.g., MacGorman et al. 2005, 2008; Wiens et al. 2005; Emersic et al. 2011). If the liquid water content is large enough, laboratory results (e.g., Takahashi and Miyawaki 2002; Saunders et al. 2006) suggest that graupel would gain a positive charge by noninductive charge exchange during rebounding collisions with cloud ice, regardless of the temperature in the mixed-phase region. As noted by MacGorman et al. (2005), the resulting charge structure would consist of a positive charge on graupel, with a negative charge on cloud ice above it and no opposite polarity of charge in lower regions. Thus, we suggest that ground flashes would not likely be produced from isolated regions of an inverted-polarity charge structure.

The environmental conditions conducive to updrafts having large enough liquid water contents to produce an inverted-polarity structure are the subject of considerable debate. All suggested hypotheses have required very strong updraft speeds over a fairly broad area, and then invoke one of the first three factors described above to limit the depletion of the adiabatic liquid water profile by precipitation below and in the mixed-phase region. Knapp (1994) and Carey and Buffalo (2007) noted that ground flashes lowering positive charge to the ground tended to occur in regions with low subcloud moisture. On this basis, Williams et al. (2005) and Carey and Buffalo (2007) suggested that the resulting high cloud bases and shallow depth of warm rain processes in these regions meant that less precipitation entered the mixed-phase region (as noted in factor 1) to deplete cloud liquid.

Note, however, that many regions of the southwestern United States also tend to have high cloud bases and less subcloud moisture, but do not have an unusually small proportion of flashes striking the ground (Boccippio et al. 2001) or an unusually large fraction of ground flashes lowering positive charge to the ground (e.g., Orville et al. 2002). It may be that another environmental characteristic such as large shear in horizontal wind with height or large values of convective available potential energy (CAPE) are also needed in combination with a high cloud base. Studies comparing lightning and storm kinematics on the high plains with those in the southwestern United States may be needed to examine the influence of high cloud base alone on ground flash production.

MacGorman et al. (2005) suggested that another reason for an enhanced probability of storms with an inverted-polarity electrical structure on the high plains is related to the tendency toward low precipitation efficiency for storms in this region. Arguments similar to those presented in factor 3 suggest that the low precipitation efficiency is due to having less recirculation of precipitation into the updraft. Again, there would then be more than the usual concentration of liquid water in the mixed-phase region because there would be less precipitation to deplete cloud liquid, perhaps leaving enough liquid water content in that region to cause an inverted-polarity electrical structure.

Two properties will affect more than one of the five factors described above. First, any mechanism for reducing precipitation between −3° and −8°C reduces electrification not only by the reduced graupel concentrations, but also by reduced secondary ice production during droplet freezing on graupel (Hallett and Mossop 1974). Because charge separation during rebounding collisions is proportional to the product of the number density of graupel with that of cloud ice, charging rates for the lower charge region will be reduced more in that temperature range than would be expected from the reduced graupel concentrations alone. Also, graupel gains positive charge as ice splinters are ejected by the Hallett–Mossop process (Hallett and Saunders 1979; Lighezzolo et al. 2010), so having less precipitation at these temperatures will also directly reduce the lower positive charge.

Second, the effect an unfavorable electrical structure in an updraft has on ground flash production can be modified by the charge structure of adjacent regions, as noted by MacGorman et al. (2005). If a lower positive charge is lacking in the strong updraft of a normal-polarity storm, it may still be supplied by noninductive charging (Takahashi and Miyawaki 2002; Saunders et al. 2006) in an adjacent region of weaker updraft, by inductive (e.g., Wilson 1929) or noninductive charging in an older adjacent cell, or by lightning depositing the charge as described by Marshall and Winn (1982). These same processes can produce a lower negative charge for strong updrafts with an inverted-polarity electrical structure (e.g., Weiss et al. 2008). For example, the lower negative charge for positive ground flashes in an inverted-polarity storm observed by Emersic et al. (2011) was provided by the descending negative region of an adjoining cell having a normal-polarity electrical structure.

Not all of the above-listed factors would be expected to affect ground flash production preferentially in the high plains. Relative to storms farther south and east, storms on the high plains tend to have a higher cloud base and a shallower layer of warm-rain processes (Williams et al. 2005; Carey and Buffalo 2007), larger storm-relative upper-level shear (e.g., Rasmussen and Straka 1998) and less recirculation of precipitation (MacGorman et al. 2005), and a greater propensity for an inverted-polarity electrical structure (Orville et al. 2002; Lang et al. 2004; Rust et al. 2005; MacGorman et al. 2005; Wiens et al. 2005; Carey and Buffalo 2007; Tessendorf et al. 2007a,b; Weiss et al. 2008). We suggest that some combination of these factors is responsible for delaying and inhibiting ground flash production in storms on the high plains through their effect on lower charge regions (i.e., below the major charge region at midlevels of storms). Additional research is needed to determine the relative contribution of each factor.

Acknowledgments

This study was supported by a Special Project Initiative grant from the NOAA/Office of Atmospheric Research. Vaisala provided the LDARII data from north Texas and the NLDN data, including the special NLDN cloud flash dataset for Oklahoma. The VHF mapping system for the high plains storms was acquired under Grants ATM-9601652 and ATM-9912073 to the New Mexico Institute of Mining and Technology from the National Science Foundation. We thank Ted Mansell and Conrad Ziegler for helpful discussions. Any opinions, findings, conclusions, or recommendations expressed in this paper are those of the authors and do not necessarily reflect the views of the National Science Foundation, Vaisala, or the NOAA/Office of Atmospheric Research.

REFERENCES

REFERENCES
Biagi
,
C. J.
,
K. L.
Cummins
,
K. E.
Kehoe
, and
E. P.
Krider
,
2007
:
National Lightning Detection Network (NLDN) performance in southern Arizona, Texas, and Oklahoma in 2003–2004
.
J. Geophys. Res.
,
112
,
D05208
,
doi:10.1029/2006JD007341
.
Boccippio
,
D. J.
,
K. L.
Cummins
,
H. J.
Christian
, and
S. J.
Goodman
,
2001
:
Combined satellite- and surface-based estimation of the intracloud-cloud-to-ground lightning ratio over the continental United States
.
Mon. Wea. Rev.
,
129
,
108
122
.
Bruning
,
E. C.
,
W. D.
Rust
,
T. J.
Schuur
,
D. R.
MacGorman
,
P. R.
Krehbiel
, and
W.
Rison
,
2007
:
Electrical and polarimetric radar observations of a multicell storm in TELEX
.
Mon. Wea. Rev.
,
135
,
2525
2544
.
Carey
,
L. D.
, and
S. A.
Rutledge
,
1996
:
A multiparameter radar case study of the microphysical and kinematic evolution of a lightning producing storm
.
J. Meteor. Atmos. Phys.
,
59
,
33
64
.
Carey
,
L. D.
, and
K. M.
Buffalo
,
2007
:
Environmental control of cloud-to-ground lightning polarity in severe storms
.
Mon. Wea. Rev.
,
135
,
1327
1353
.
Carey
,
L. D.
,
M. J.
Murphy
,
T. L.
McCormick
, and
N. W. S.
Demetriades
,
2005
:
Lightning location relative to storm structure in a leading-line, trailing-stratiform mesoscale convective system
.
J. Geophys. Res.
,
110
,
D03105
,
doi:10.1029/2003JD004371
.
Coleman
,
L. M.
,
M.
Stolzenburg
,
T. C.
Marshall
, and
M.
Stanley
,
2008
:
Horizontal lightning propagation, preliminary breakdown, and electric potential in New Mexico thunderstorms
.
J. Geophys. Res.
,
113
,
D09208
,
doi:10.1029/2007JD009459
.
Cummins
,
K. L.
, and
M. J.
Murphy
,
2009
:
An overview of lightning locating systems: History, techniques, and data uses, with an in-depth look at the U.S. NLDN
.
IEEE Trans. Electromagn. Compat.
,
51
,
499
518
.
Cummins
,
K. L.
,
M. J.
Murphy
,
E. A.
Bardo
,
W. L.
Hiscox
,
R. B.
Pyle
, and
A. E.
Pifer
,
1998
:
A combined TOA/MDF technology upgrade of the U.S. National Lightning Detection Network
.
J. Geophys. Res.
,
103
,
9035
9044
.
Emersic
,
C.
, and
C. P. R.
Saunders
,
2010
:
Further laboratory investigations into the Relative Diffusional Growth Rate theory of thunderstorm electrification
.
Atmos. Res.
,
98
,
327
340
,
doi:10.1016/j.atmosres.2010.07.011
.
Emersic
,
C.
,
P. L.
Heinselman
,
D. R.
MacGorman
, and
E. C.
Bruning
,
2011
:
Lightning activity in a hail-producing storm observed with phased-array radar
.
Mon. Wea. Rev.
,
138
,
1809
1825
.
Fleenor
,
S. A.
,
C. J.
Biagi
,
K. L.
Cummins
,
E. P.
Krider
, and
W.-M.
Shao
,
2009
:
Characteristics of cloud-to-ground lightning in warm-season thunderstorms in the central Great Plains
.
Atmos. Res.
,
91
,
333
352
,
doi:10.1016/j.atmosres.2008.08.011
.
Goodman
,
S. J.
,
D. E.
Buechler
,
P. D.
Wright
, and
W. D.
Rust
,
1988
:
Lightning and precipitation history of a microburst-producing storm
.
Geophys. Res. Lett.
,
15
,
1185
1188
.
Hallett
,
J.
, and
S. C.
Mossop
,
1974
:
Production of secondary ice crystals during the riming process
.
Nature
,
249
,
26
28
.
Hallett
,
J.
, and
C. P. R.
Saunders
,
1979
:
Charge separation associated with secondary ice
.
J. Atmos. Sci.
,
36
,
2230
2235
.
Jacobson
,
E. A.
, and
E. P.
Krider
,
1976
:
Electrostatic field changes produced by Florida lightning
.
J. Atmos. Sci.
,
33
,
103
117
.
Johnson
,
E. V.
, and
E. R.
Mansell
,
2006
:
Three-dimensional lightning mapping of the central Oklahoma supercell on 26 May 2004
.
Preprints, Second Conf. on Meteorological Applications of Lightning Data, Atlanta, GA, Amer. Meteor. Soc., 6.5. [Available online at http://ams.confex.com/ams/Annual2006/techprogram/paper_104352.htm.]
Knapp
,
D. I.
,
1994
:
Using cloud-to-ground lightning data to identify tornadic thunderstorm signatures and nowcast severe weather
.
Natl. Wea. Assoc. Dig.
,
19
,
35
42
.
Krehbiel
,
P. R.
,
1986
:
The electrical structure of thunderstorms
.
The Earth’s Electrical Environment, National Academies Press, 90–113
.
Kuhlman
,
K. M.
,
C. L.
Ziegler
,
E. R.
Mansell
,
D. R.
MacGorman
, and
J. M.
Straka
,
2006
:
Numerically simulated electrification and lightning of the 29 June 2000 STEPS supercell storm
.
Mon. Wea. Rev.
,
134
,
2734
2757
.
Lang
,
T. J.
, and
Coauthors
,
2004
:
The Severe Thunderstorm Electrification and Precipitation Study
.
Bull. Amer. Meteor. Soc.
,
85
,
1107
1125
.
Larson
,
H. R.
, and
E. J.
Stansbury
,
1974
:
Association of lightning flashes with precipitation cores extending to height 7 km
.
J. Atmos. Terr. Phys.
,
36
,
1547
1553
.
Lhermitte
,
R.
, and
P. R.
Krehbiel
,
1979
:
Doppler radar and radio observations of thunderstorms
.
IEEE Trans. Geosci. Electron.
,
GE-17
,
162
171
.
Lighezzolo
,
R. A.
,
R. G.
Pereyra
, and
E. E.
Avila
,
2010
:
Measurements of electric charge separated during the formation of rime by the accretion of supercooled droplets
.
Atmos. Chem. Phys.
,
10
,
1661
1669
.
Lund
,
N.
,
D.
MacGorman
,
T.
Schuur
,
M.
Biggerstaff
, and
D.
Rust
,
2009
:
Relationships between lightning location and polarimetric radar signatures in a small mesoscale convective system
.
Mon. Wea. Rev.
,
137
,
4151
4170
.
MacGorman
,
D. R.
, and
W. D.
Rust
,
1998
:
The Electrical Nature of Storms
.
Oxford University Press, 422 pp
.
MacGorman
,
D. R.
,
D. W.
Burgess
,
V.
Mazur
,
W. D.
Rust
,
W. L.
Taylor
, and
B. C.
Johnson
,
1989
:
Lightning rates relative to tornadic storm evolution on 22 May 1981
.
J. Atmos. Sci.
,
46
,
221
250
.
MacGorman
,
D. R.
,
J. M.
Straka
, and
C. L.
Ziegler
,
2001
:
A lightning parameterization for numerical cloud models
.
J. Appl. Meteor.
,
40
,
459
478
.
MacGorman
,
D. R.
,
W. D.
Rust
,
P.
Krehbiel
,
W.
Rison
,
E.
Bruning
, and
K.
Wiens
,
2005
:
The electrical structure of two supercell storms during STEPS
.
Mon. Wea. Rev.
,
133
,
2583
2607
.
MacGorman
,
D. R.
,
T.
Filiaggi
,
R. L.
Holle
, and
R. A.
Brown
,
2007
:
Negative cloud-to-ground lightning flash rates relative to VIL, maximum reflectivity, cell height, and cell isolation
.
J. Lightning Res.
,
1
,
132
147
.
MacGorman
,
D. R.
, and
Coauthors
,
2008
:
TELEX: The Thunderstorm Electrification and Lightning Experiment
.
Bull. Amer. Meteor. Soc.
,
89
,
997
1013
.
Mansell
,
E. R.
,
D. R.
MacGorman
,
J. M.
Straka
, and
C. L.
Ziegler
,
2002
:
Simulated three-dimensional branched lightning in a numerical thunderstorm model
.
J. Geophys. Res.
,
107
,
4075
,
doi:10.1029/2000JD000244
.
Mansell
,
E. R.
,
D. R.
MacGorman
,
C. L.
Ziegler
, and
J. M.
Straka
,
2005
:
Charge structure in a simulated multicell thunderstorm
.
J. Geophys. Res.
,
110
,
D12101
,
doi:10.1029/2004JD005287
.
Mansell
,
E. R.
,
C. L.
Ziegler
, and
E. C.
Bruning
,
2010
:
Simulated electrification of a small thunderstorm with two-moment bulk microphysics
.
J. Atmos. Sci.
,
67
,
171
174
.
Marshall
,
T. C.
, and
W. P.
Winn
,
1982
:
Measurements of charged precipitation in a New Mexico thunderstorm: Lower positive charge centers
.
J. Geophys. Res.
,
87
,
7141
7157
.
Mazur
,
V.
, and
L. H.
Ruhnke
,
1998
:
Model of electric charges in thunderstorms and associated lightning
.
J. Geophys. Res.
,
103
,
23 299
23 308
.
Orville
,
R. E.
,
2008
:
Development of the National Lightning Detection Network
.
Bull. Amer. Meteor. Soc.
,
89
,
180
190
.
Orville
,
R. E.
,
G. R.
Huffines
,
W. R.
Burrows
,
R. L.
Holle
, and
K. L.
Cummins
,
2002
:
The North American Lightning Detection Network (NALDN)—First results: 1998–2000
.
Mon. Wea. Rev.
,
130
,
2098
2109
.
Payne
,
C. D.
,
T. J.
Schuur
,
D. R.
MacGorman
,
M. I.
Biggerstaff
,
K. M.
Kuhlman
, and
W. D.
Rust
,
2010
:
Polarimetric and electrical characteristics of a lightning ring in a supercell storm
.
Mon. Wea. Rev.
,
138
,
2405
2425
.
Rasmussen
,
E. N.
, and
J. M.
Straka
,
1998
:
Variations in supercell morphology. Part I: Observations of the role of upper-level storm-relative flow
.
Mon. Wea. Rev.
,
126
,
2406
2421
.
Reap
,
R. M.
, and
D. R.
MacGorman
,
1989
:
Cloud-to-ground lightning: Climatological characteristics and relationship to model fields, radar observations, and severe local storms
.
Mon. Wea. Rev.
,
117
,
518
535
.
Rison
,
W.
,
R. J.
Thomas
,
P. R.
Krehbiel
,
T.
Hamlin
, and
J.
Harlin
,
1999
:
A GPS-based three dimensional lightning mapping system: Initial observations in New Mexico
.
Geophys. Res. Lett.
,
26
,
3573
3576
.
Rust
,
W. D.
, and
D. R.
MacGorman
,
2002
:
Possible inverted-polarity electrical structures in thunderstorms during STEPS
.
Geophys. Res. Lett.
,
29
,
1571
,
doi:10.1029/2001GL014303
.
Rust
,
W. D.
, and
Coauthors
,
2005
:
Inverted-polarity electrical structures in thunderstorms in the Severe Thunderstorm Electrification and Precipitation Study (STEPS)
.
Atmos. Res.
,
76
,
247
271
,
doi:10.1016/j.atmosres.2004.11.029
.
Saunders
,
C. P. R.
,
H.
Bax-Norman
,
C.
Emersic
,
E. E.
Avila
, and
N. E.
Castellano
,
2006
:
Laboratory studies of the effect of cloud conditions on graupel/crystal charge transfer in thunderstorm electrification
.
Quart. J. Roy. Meteor. Soc.
,
132
,
2653
2673
.
Schultz
,
C. J.
,
W. A.
Petersen
, and
L. D.
Carey
,
2009
:
Preliminary development and evaluation of lightning jump algorithms for the real-time detection of severe weather
.
J. Appl. Meteor. Climatol.
,
48
,
2543
2563
.
Takahashi
,
T.
, and
K.
Miyawaki
,
2002
:
Reexamination of riming electrification in a wind tunnel
.
J. Atmos. Sci.
,
59
,
1018
1025
.
Tessendorf
,
S. A.
,
K. C.
Wiens
, and
S. A.
Rutledge
,
2007a
:
Radar and lightning observations of the 3 June 2000 electrically inverted storm from STEPS
.
Mon. Wea. Rev.
,
135
,
3665
3681
.
Tessendorf
,
S. A.
,
S. A.
Rutledge
, and
K. C.
Wiens
,
2007b
:
Radar and lightning observations of normal and inverted polarity storms from STEPS
.
Mon. Wea. Rev.
,
135
,
3682
3706
.
Thomas
,
R.
,
P.
Krehbiel
,
W.
Rison
,
S.
Hunyady
,
W.
Winn
,
T.
Hamlin
, and
J.
Harlin
,
2004
:
Accuracy of the lightning mapping array
.
J. Geophys. Res.
,
109
,
D14207
,
doi:10.1029/2004JD004549
.
Weiss
,
S.
,
W. D.
Rust
,
D. R.
MacGorman
,
E.
Bruning
, and
P.
Krehbiel
,
2008
:
Evolving complex electrical structure of the STEPS 25 June 2000 multicell storm
.
Mon. Wea. Rev.
,
136
,
741
756
.
Wiens
,
K. C.
,
S. A.
Rutledge
, and
S. A.
Tessendorf
,
2005
:
The 20 June 2000 supercell observed during STEPS. Part II: Lightning and charge structure
.
J. Atmos. Sci.
,
62
,
4151
4177
.
Williams
,
E.
, and
Coauthors
,
1999
:
The behavior of total lightning activity in severe Florida thunderstorms
.
Atmos. Res.
,
51
,
245
265
.
Williams
,
E.
,
V.
Mushtak
,
D.
Rosenfeld
,
S.
Goodman
, and
D.
Boccippio
,
2005
:
Thermodynamic conditions favorable to superlative thunderstorm updraft, mixed phase microphysics, and lightning flash rate
.
Atmos. Res.
,
76
,
288
306
,
doi:10.1016/j.atmosres.2004.11.009
.
Wilson
,
C. T. R.
,
1929
:
Some thundercloud problems
.
J. Franklin Inst.
,
208
,
1
12
.
Young
,
K. C.
,
1993
:
Microphysical Processes in Clouds
.
Oxford University Press, 427 pp
.
Ziegler
,
C. L.
,
E. R.
Mansell
, and
E. C.
Bruning
,
2010
:
Impact of varying CCN concentration on the precipitation process of a simulated convective storm
.
Preprints, 13th Conf. on Cloud Physics, Portland, OR, Amer. Meteor. Soc., JP3.17. [Available online at http://ams.confex.com/ams/pdfpapers/171866.pdf.]