RELATING CLOUD-TO-GROUND LIGHTNING TO SEVERE WEATHER IN INDIANA ON 2 JUNE 1990

Transcription

1 RELATING CLOUD-TO-GROUND LIGHTNING TO SEVERE WEATHER IN INDIANA ON 2 JUNE 199 David B. Elson National Weather Service Forecast Office Indianapolis, Indiana Abstract In this study, c1oud-to-ground lightning activity is statistically compared to severe eather occurrences ith thunderstorms in Indiana on 2 June 199. Emphasis is placed on rates and trends of the lightning activity in these storms. Five-minute lightning rates versus time ere plotted for each of 14 different" storms. Severe eather occurrences (tornadoes, high inds and large hail) ere then assigned to one of these storms (hen possible). Statistical tests ere run on the data to measure correlations beteen lightning rates and the three types of severe eather. Also, the locations of severe eather ith respect to the approximate centroid of lightning activity ere plotted. Of particular interest as the strong tendency of high (straight line) inds to occur during or shortly after a period of high lightning activity. Tornadoes, on the other hand, shoed a fairly strong tendency to occur during periods of lo to moder(lte lightning activity. For all types of severe eather combined, the centroid of lightning activity exhibited a tendency to occur a fe miles to the south and east of the severe eather. 1. Introduction In the past decade, an increasing research effort has attempted to correlate lightning activity ith severe eather. Unfortunately, these studies have found conflicting results (Kane 1991). The general consensus from these studies indicates that c1oud-to-ground lightning rates appear to have little predictive value for tornadoes (MacGorman et at. 1989; Maier and Krider 1982). Both MacGorman et at. (1989) and Maier and Krider (1982) noted in fact, that lightning activity actually increased at the end oftornadic activity. Conversely, in studies based on satellite observations, Turman and Tettelbach (198) found lightning activity to be more frequent in tornadic storms. Past studies indicate that lightning rates hold more potential as a precursor to non-tornadic severe eather (large hail and high inds). Maier and Krider (1982), in comparing severe thunderstorms in Oklahoma to thunderstorms in Florida, found both mean and peak lightning rates in severe thunderstorms to be to or more times higher than the rates observed in non-severe thunderstorms. Goodman and MacGorman (1986) in studying lightning rates in mesoscale convective complexes (MCC) (Maddox 198) found 43 percent of all severe eather occurrences associated ith MCCs to occur during the time of the greatest increase in lightning rates. This study ill focus on the relationship beteen lightning intensity and severe eather occurrences during the 2 June 199 tornado outbreak over Indiana. All storms developed along a single pre-frontal ave, except for one, hich folloed in the others ake by about one hour. The five-minute hourly rates (a five minute stroke count expressed as an hourly rate) of all cloud-to-ground lightning activity ill be compared to severe eather occurrences for fourteen thunderstorms. No attempt is made in this study to differentiate beteen positive and negative lightning flashes. 2. Methodology Fourteen storms (Fig. 1) ere identified using lightning data obtained from the SUNYA (State University of Ne York at Albany) lightning detection netork (Orville et at. 1983). Only storms occurring in the state of Indiana during the period from 18 to 24 EST2 June 199 ere examined. Lightning data and severe eather occurrences after 2125 EST in part of southeast Indiana ere omitted from this study (see Figs. I and 2). The same basic technique as Kane (1991) as used to examine the data, in hich five-minute hourly rates in lightning activity ere compared ith severe Fig. 1. Approximate storm tracks for lightning activity centroids in Indiana 18 to 24 EST 2 June 199. Shaded region is area of omitted data. 15

2 16 eather occurrences (Storm Data, June 199) (Fig. 2). Other data sources (i.e., radar) ere not utilized in this study. Fourteen individual lightning maxima ere identified. Although every effort as made to keep the lightning count areas as small and equally sized as possible, no set size for a count area as utilized. Instead, the count areas ere adjusted for each storm continuously to best fit the storm size and location. Although no measurements of size of the count areas ere made, the average size is estimated to be 6 square miles. It should be noted that by varying the size of count areas to accommodate for the fluid nature of the storms, the intensity is in terms of lightning activity for the storm as a hole, and should not be confused ith intensity in terms of lightning activity per unit area. The terms used to describe lightning activity are relative to the other storms obse\ ved across Indiana that day. When c1oud-to-ground lightning activity exceeded a rate of 3 strikes per hour, it as classified as a period of high lightning activity. Hourly rates of 1 to 3 strikes ere classified as moderate activity, and storms ith less than 1 strikes per hour ere classified as lo or eak lightning activity. Binomial distributions ere the chief statistical test utilized for analysis of the data. In using binomial distributions as a test, it as necessary to first assume that severe eather National Weather Digest as completely independent of lightning activity. By comparing the total amount of time and/or number of occurrences of severe eather ith the amount of time and/or number of occurrences during a selected period of interest, a probability for such a distribution can be computed. If the probability of such a distribution as :s 5 percent, it as termed significant. 3. Individual Storm Characteristics A generalized summary for individual storm characteristics is contained in Table I. Folloing are other significant characteristics of the storms. Storm I had the largest sustained output of lightning of all storms analyzed. For over one hour, the storm averaged an hourly lightning rate beteen 4 and 55 lightning strikes, during hich the three reports of high inds occurred. Storm I, along ith storms 2 through 4 tended to exhibit the most diffuse lightning strike patterns, often apparently being connected to adjac.ent lightning maxima. Storm 2 (Fig. 3) produced the highest five minute lightning count of all storms analyzed, ith a peak five-minute hourly rate exceeding 7 strikes. During the course of an hour and a half, this storm rivaled storm I for sustained lightning production, but shoed large slo-changing lightning rates. The to reports of severe thunderstorm inds from storm 2 folloed the final major collapse in lightning production by 15 to 3 minutes. The one report of severe thunderstorm inds ith storm 3 in Indiana occurred near the Illinois state line at 18 EST. Because lightning data ere examined only after 18 EST, this ind event as not considered in the statistical analyses. Storm 4 folloed very closely the track of storm 3, hich preceded storm 4 by about fifty minutes. Storm 4 exhibited the most diffuse pattern of lightning strikes. This coupled ith storm 4's movement near storm 3's track, made it the most difficult storm from hich to obtain accurate counts. The ind and hail reports ere scattered throughout the storm' s lifetime in Indiana. Four of the ind events and all three hail events occurred ithin 2 minutes of lightning rate peaks exceeding 27 strokes per hour. One of the tornadoes occurred during lo lightning activity just 1 minutes after initial storm development. The second occuri'ed just after a local peak in moderate lightning activity about halfay through the storm's life. Fig. 2. Severe eather occurrences in Indiana 18 to 24 EST 2 June 199. This includes separate reports counted as one occurrence. "A" indicates large hail, "W" indicates ind damage, "T" and lines indicate Tornadoes. Shaded region is area of omitted data. Table 1. Individual storm characteristics. Predominant High Lightning Large Hail Wind Storm # Activity Tornadoes Reports Reports 1 High None None 3 2 High None None 2 3 Lo None None 1 4 Moderate Lo 6 None None 5A Moderate 1 None None 6 Large Variability Lo None None None 8 Lo 2 None None 9 Moderate 7 None None 9A Lo 2 None None 1 Moderate 1 None None 11 Lo 2 None None 12 Moderate None None None

3 Volume 18 Number 3 December, ~ ~ ~ WW L 7 :J I "'-... (f) OJ 5 :::f. '\ L -(f) '-.-/ 4 OJ - :::: (J)3 c c - ~ 2 ~ 1 o ~-.-.~'-'-r-r-r-.-.-'-' rl-'i-'i-'i-'i-'i-'i-'i'-'-r-r-r-.-.-'-' ~ Time (EST) Fig. 3. Cloud-to-ground lightning rate vs. time for storm 2. "W" indicates time of occurrence of ind damage. Storm 5 only briefly exceeded an hourly rate of 1 strikes, but produced six tornadoes, three of them FOs. The to FI tornadoes and the one F2 shoed a tendency to occur during periods of lo lightning activity. Storm SA appears to have developed along the southern flank of storm 5. This storm produced a short-lived FI tornado at a time of relatively lo lightning activity. Storm 6 initially, foroveran hour, as a moderate lightning producer. During the next hour it as a prolific producer, then finished as a moderate to eak producer. At peak activity (early in the prolific stage) this storm had a five-minute hourly rate of lightning strikes near 6. Seven of the eight tornadoes occurred during the first half of the storms lifetime in Indiana, during the time period preceding peak lightning activity. Four of the tornadoes occurred in the tenty minute transition stage going from moderate lightning production to prolific production, hich ould tend to support the findi ngs of MacGorman et aj. (1989) and Maier and Krider (1982). Only one small tornado occurred after the peak in lightning activity. The to reports ofjarge hail occurred roughly fifteen minutes after the drop-off of to separate lightning activity peaks. The three high ind events occurred during and after the peak lightning activity. Although storm 8 as the most feeble lightning producer of the day, the storm did produce to tornadoes in its brief lifetime, an FO and an F2. Storm 9 (Fig. 4) as a moderate lightning producer for about half its lifetime, and a eak lightning producer over. the other half. Its peak activity reached a five-minute hourly strike rate of36 strikes. This storm produced three F4 tornadoes, one F3, to F2s and one FO. Tornadoes ere on the ground for roughl y to-thirds of this storm's life in Indiana. The tornadoes shoed a preference to form during periods of relatively lo lightning activity. Storm 9A appeared to split off from storm 9. Within the first hour after initial identification, this storm produced to small tornadoes. Storm IO as analyzed for only the first hour and a half of its life in Indiana. The second half of this storm's life, plus another separate storm in southeast Indiana ere not analyzed (Figs. I and 2) (these unanalyzed storms accounted for three more tornadoes and one high ind occurrence hich ere not included in the statistical analysis). Storm IO as credited ith spaning one F4 tornado. This storm as the closest detectable lightning maxima to an additional three tornadoes, including an F4 hich had a damage track

4 18 National Weather Digest _-- -_._._._ ,,--.,. \... :::J 6 I '-...,. (J) F3 Q) F4 F2 ~ \... +-, U1 FO F4 F2 F4 '--./ Q) +- ::: (J) 3 t: c +- ~ O~~~~~~~~~,,~~~-~4rrr~.T+.~~~~~~~~ Time (EST) Fig. 4. Cloud-to-ground lightning rate vs. time for storm 9. " F#" indicates time of occurrence of tornadoes, rated on the Fujita scale. of over 9 miles. These additional tornadoes ere not credited to storm 1 since the reports ere 2 miles or more ahead of the lightning maxima. For statistical purposes later, these additional reports ere counted as unrelateable to the lightning data. Storm II as a eak lightning producer for the first to hours, and a moderate lightning producer for the folloing hour. The to tornadoes both occurred during the period of moderate lightning production. 4. Analysis and Discussion a. Introduction and methodology The 2 June 199 tornado outbreak as the largest in Indiana history. Thirty-seven tornado touchdons ere confirmed in Indiana, most occurring across the southern half of the state. The storms in the northern half of the state ere more noted for damaging inds. The fourteen storms analyzed accounted for a total of 1,91 storm-lightning minutes (Table 2). Because of the high frequency of severe eather associated ith these storms, and because data ere collected for only one day, one must be very careful about draing any conclusions on relationships beteen lightning frequency and severe eather. Holle and Maier (1982) found that the greatest rate of lightning is associated ith the echoes having the greatest vertical development. Extrapolating from that, Goodman et al. (1988) suggest higher lightning rates may be the result of a stronger updraft. Nielson and MacGorman (1989) suggest that the updraft is important in the generation of charge. Additionally, one might conclude that a decrease in lightning activity folloing a local peak in activity ould most likely be a time for the development of a strong dondraft. This i~ supported by Goodman et al. (1988), ho found that the decrease in flash rate associated ith storm collapse also served as a precursor to the arrival of a microburst at the surface. Table 2. Time (minutes) Percent of Total Lightning-minutes breakdon. All Storms Lightning Rate 2: Lightning Rate 2: 3 plus 6 min after

5 1 Volume 18 Number 3 December, 1993 To support these claims, the number of minutes hen lightning activity equaled or exceeded an hourly rate of 3 strikes, plus the sixty minutes immediately folloing each occurrence ere counted. This as done chiefly to analyze the relationship beteen damaging inds and peak lightning activity, since one donburst may initiate a gust front at the surface lasting a number of minutes after the donburst occurrence. In this category, 494 storm-lightning minutes ere counted for all storms, hich equaled approximately 26 percent of the total storm-lightning time observed. The second method of comparison beteen severe eather and c1oud-to-ground lightning rates counted only the number of minutes hen lightning activity equaled or exceeded an hourly rate of 3 strikes. This as done to analyze any direct relationships beteen concurrent lightning activity and se\\ere eather. Total storm time hen the lightning activity rates equaled or exceeded an hourly rate of 3 strikes as 228 minutes, or 12 percent of total stormlightning time. The threshold rate of 3 strikes per hour as chosen as best describing the periods of highest c1oud-to-ground lightning activity, as opposed to periods of lesser activity for this data set. This value may be found to change ith differing storm environments. The final association investigated beteen severe eather and c1oud-to-ground lightning activity as spacial relationships. The location of all severe eather occurrences ith respect to the approximate centroid of lightning activity as plotted. b. Time relationships Results of the statistical tests are summarized in Table 3. Note that the total number of occurrences of hail and tornadoes in Tables I and 3 do not match since not all occurrences ere able to be matched ith a storm. Of the thirty-four tornadoes analyzed, thirty-three (or 97 percent) ere found to occur during periods hen lightning activity as belo 3 strikes per hour. Only one tornado (associated ith storm 6) occurred during a period above that lightning rate. If it is assumed that there is an equal chance for tornadoes to occur at any time during a storm, then the probability for one or less tornadoes to occur during a period of high lightning activity (rate > 3 strikes/hour) is only 4.8 percent for this case. This is significant enough to state that based on this data set, tornadoes are much more likely to occur during periods of lo or moderate lightning activity. For the sake of completeness, a similar comparison of the five tornadoes observed during or less than 6 minutes after high lightning activity as made, and found to be just slightly less significant. The probability for five or feer tornadoes during that period is 5.2 percent. Of sixteen analyzed instances of ind damage, telve ere found to occur either during or ithin sixty minutes after a period of high lightning activity. Assuming an equal probability for ind damage to occur at any time during a storm, the probability for telve or more of the ind damage occurrences to happen during the aforementioned period is only.6 percent. This is quite significant. Based on the analysis of this data set, it appears much more likely for ind damage to occur either during or ithin sixty minutes after a period of high lightning activity. This is in agreement ith Goodman et al. (1988), shoing that a decrease in flash rate serves as a precursor to the arrival of a microburst. Table 3. Type of Severe Weather Tornado High Winds Large Hail Time of occurrence relationships. No. of occurrences during high lightning Total No. of activity Analyzed No. of Significance Occurrences occurrences during or ithin 6 mins after high lightning activity Tornadoes most 1 likely to occur during periods of 5 lo to moderate lightning activity o 3 19 High inds much more likely to occur during or after high lightning activity No Significance A similar analysis comparing ind damage to periods of high lightning activity as not quite as conclusive. The probability of the four instances of ind damage occurring during the period of high lightning activity is 12 percent. There ere seven reports oflarge hail (::::: ~ inch diameter), but only five ere assigned to storms. These five reports ere accounted for by to storms. Because of the small size of the hail data set, no conclusions could be dran. The hail statistics are included here only for completeness. None of the ha il as reported during a time of high lightning activity. This represents a probability of 41 percent. Hoever, three of the hail reports ere ithin sixty minutes after a period of high lightning activity. The probability of three or more occurrences of large hail during this period is 26 percent. In considering the results of the statistical tests presented here, a fe considerations need to be made. First, the occurrences of severe eather ere assumed to be independent of each other. This obviously is not alays the case, but because National Weather Service verification practices count each county as an independent occurrence, and since there is no consistently accurate method to differentiate beteen dependent and independent occurrences, the National Weather Service practice as extended here. Second, it is assumed that all severe eather occurrences are knon, and their time of occurrence is accurate. With respect to other lightning trends (increasing or decreasing activity, local peaks or lulls, etc.) at times of severe eather, no other correlations ere apparent ith this case. c. Spacial relationships A plot of the location of all severe eather occurrences ith respect to the approximate centroid of lightning activity (Fig. 5) reveals a ide variability in locations of severe

6 2 National Weather Digest N T E Miles 21 Miles Fig. 5. Plot of the storm relative location of severe eather occurrences ith respect to the approximate lightning centroid. "A" indicates large hail, " W" indicates ind damage, "r' and lines indicate tornadoes. eather. Some of this variability may be accounted for by the subjective nature in determining the lighting centroids. Despite the limitations of subjectivity, it is apparent that the most densely covered area (for all types of severe eather) is a fe miles to the est of the centroid. Very fe occurrences of severe eather ere noted in the southeast quadrant of lightning activity. 5. Summary The intent of this study as to investigate relations beteen cloud-to-ground lightning activity and the occurrence of severe eather ith storms over Indiana on 2 June 199. Damaging inds shoed a strong tendency to occur during, or ithin sixty minutes after a period of high lightning activity. This is similar to the findings by Goodman et al. (1988) in hich peak lightning activity as found to occur approximately ten minutes prior to the arrival of the microburst at the surface. High inds ere not exclusive to these periods hoever, as four of the sixteen instances of ind damage ere noted to occur outside of this time frame. Tornadoes shoed a tendency to occur during periods of lo to moderate lightning activity, in agreement ith MacGorman et al. (1989) and Maier and Krider (1982). Thirty-three of the thirty-four tornadoes occurred during lo to moderate lightning activity. Hoever, it should be noted that lo to moderate lightning activity as prevalent ith the storms analyzed, accounting for 88 percent of all storm time. Insufficient data ere collected for analysis of hail events. Because of the nature of the data set analyzed (an abnormally high frequency of severe eather occurrences), no conclusions could be dran as to the probability of severe eather relative to cloud-to-ground lightning activity. Severe eather tended to not occur to the southeast of the estimated center of lightning activity. The preferred area for severe eather appeared to be a fe miles est of this center, although a large amount of variability as noted. Based on this study, cloud-to-ground lightning activity) does seem to hold some potential as a predictor of severe eather (particularly for strong inds), hich ould arrant further investigation. Lightning activity of both severe and non-severe storms needs to be compared, suggesting a need for a similar study, but on a much larger scale. Acknoledgments I ould like to thank the staff of the National Weather Service Forecast Office (NWSFO) at Indianapolis for their help and comments, particularly Bill Gery, John Hendrick-

7 " Volume 18 Number 3 December, 1993 son, and Mike Sabones. I ould also like to thank Tom Magnuson (NWSFO MSP), Ron Przybylinski (NWSFO STL), Richard McNulty (NWS Training Center) and Ron Holle (NSSL) for their revies at various stages, and Dan Carr (George Mason University) and Mark Delisi (NWSFO WBC) for help ith the statistics. Finally, I ish to thank the people at PSI Energy for alloing access to their archived data, particularly Brian Gunn for his help and patience. Author David Elson received his Bachelor of Science Degree in Meteorology from the North Carolina State University in Raleigh; December He subsequently orked for the Genesis of Atlantic Los Experiment (GALE) early in In May of 1986, ht!\began his career in the National Weather Service. He served first as a Meteorologist Intern in Scottsbluff, Nebraska and Fort Wayne, Indiana, then accepted a position as Hydrologist at the River Forecast Center in Harrisburg, Pennsylvania. Since March of 1991, he has been a Forecaster at the National Weather Service Forecast Office in Indianapolis, Indiana. References Goodman, S. J. and D. R. MacGorman, 1986: "Cloud to Ground Lightning Activity in Mesoscale Convective Complexes", Mon. Wea. Rev., 18, , D. E. Buecheler, P. D. Wright, and W. D. Rust, 1988: "Lightning and Precipitation History ofa Microburst-Producing Storm", Geaphys. Res. Lett., 15, Holle, R. L. and M. W. Maier, 1982: "Radar Echo Height Related to Cloud-Ground Lightning in South Florida", Preprints: 12th Conference on Severe Local Storms, San Antonio, Texas, Amer. Meteor. Soc., Kane, R. 1., 1991: "Correlating Lightning to Severe Local Storms in the Northeastern United States", Wea. Forecasting, 6,3-12. 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 ", 1. Atmas. Sci., 46, Maddo,x, R. A., 198: " Mesoscale Convective Complexes", BII(f. Amer. Meteor. Soc., 61, Maier, M. W. and E. P. Krider, 1982: " A Comparative Study of Cloud-to-Ground Lightning Characteristics in Florida and Oklahoma Thunderstorms", Preprints: Telfth Conference on Severe Local Storms, San Antonio, Texas, Amer. Meteor. Soc., Nielson, K. E. and D. R. MacGorman, 1989: "Lightning Ground Flash Rates Relative to Mesocyclone Evolution on 8 May 1986", Preprints: Tenty-jollrth Conference on Radar Meteorology, Tallahassee, Florida, Amer. Meteor. Soc., Orville, R. E., R. W. Henderson and L. F. Bosart, 1983: " An East Coast Lightning Detection Netork", BIlII. Amer. Meteor. Soc., 64, Turman, B. N. and R. J. Tuttelbach, 198: "Synoptic Scale Satellite Lightning Observations in Conjunction ith Tornadoes", Mon. Wea. Rev., 18,