Cloud-to-ground lightning data for the years 1992–95 have been analyzed for geographical distribution of total flashes, positive flashes, and the percentage of flashes that lower positive charge to ground. In the contiguous United States the measured total cloud-to-ground lightning flash counts were 16.3 million (1992), 24.2 million (in both 1993 and 1994), and 22.3 million in 1995. The maximum flash densities occurred in Florida in 1992 (9–11 flashes per square kilometer) and in the Midwest in 1993 (11–13 flashes per square kilometer), coinciding with the storms and floods that dominated the summer of 1993 in the Midwest. In 1994, the area of maximum flash density was again in Florida (11–13 flashes per square kilometer). In 1995, the flash density maxima (9–11 km−2) were in southern Louisiana and near the Kentucky–Illinois border. Positive flash densities had maxima in the Midwest in all four years with values of 0.4 (1992), 1.0 (1993), 0.7 (1994), and 1.8 flashes per square kilometer (1995). The annual mean percentage of flashes that lowered positive charge to ground was between 4% and 5% for the three years, 1992–94, but increased to 9.3% in 1995. The monthly values of the percentage of positive flashes ranged from 3% (August 1992) to 25% (December 1993). The positive flash maxima in the Midwest appear to be near the geographical areas in which cloud-ionosphere discharges (sprites) have been reported.
Cloud-to-ground lightning flash density results for the contiguous United States have been published by Orville (1991, 1994) for the years 1989–91. In the present paper, we extend these results four years to cover the most recently acquired cloud-to-ground lightning information for the period 1992–95.
Flash density measurements, or the number of lightning flashes to ground per unit area per unit time, are of fundamental interest on all spatial and temporal scales. Contour plots of the annual cloud-to-ground lightning flash counts reveal the areas with the most frequent lightning from year to year. We will see that this varies from Florida (1992 and 1994), to the St. Louis area (1993), to the Kentucky–Illinois border and southern Louisiana (1995). Recordings of cloud-to-ground positive lightning show a numerical increase over the four years, from 0.6 million (1992) to 2.1 million (1995). The percentage of positive lightning, however, remains between 4% and 5% (1992–94), increasing to 9.3% in 1995. The data, results, and discussion of these 1992–95 measurements are presented in the following sections.
All data were obtained by the National Lightning Detection Network, which is operated by the GeoMet Data Services, Inc. (now Global Atmospherics, Inc.) Tucson, Arizona. The network consists of over 100 wideband magnetic direction finders (Orville 1991, 1994) augmented by time-of-arrival (TOA) sensors beginning in July 1994 (see Fig. 1). It is believed and assumed that augmentation of the TOA sensors does not significantly increase the number of flashes recorded. This assumption is supported by the data presented in Table 1 and Fig. 2, which show that the number of flashes recorded since July 1994 appears to vary little month by month when compared to a previous year such as 1993. This comparison is not true, however, for the year 1992 when there was generally less lightning on a month-by-month basis compared to subsequent years.
Locations of the magnetic direction finders in the contiguous United States have been published previously (Orville 1991, Fig. 2; Orville 1994, Fig. 1). Changes in the network configuration in subsequent years (1994–95) have resulted in a modified configuration shown in Fig. 1. We assume that the modified configuration does not affect significantly the results in this paper. The principles of magnetic direction finding, detection efficiency, and location errors have been discussed previously by several authors (e.g., Krider et al. 1976,1980; Mach et al. 1986; Orville 1994) and will not be repeated here. In our present analysis, we assume that a 70% detection efficiency applies throughout the network. Recent detection efficiency measurements in New York reported by Pyle (1995) show an average efficiency of 76%. All flash density results in this paper are the result of multiplying the measurements by a factor of 1.4 to correct for the 70% detection efficiency. It is not obvious, however, that the detection efficiency is the same for negative and positive flashes. Hojo et al. (1989, Fig. 1) using a magnetic direction finder network in the mid-1980s report a detection efficiency for positive flashes that is less than for negative flashes at all distances from their medium gain direction finders. In the absence of separate measurements of negative and positive flash detection efficiencies for the advanced lightning direction finders (ALDF), we will assume that the detection efficiencies are the same. Location errors are assumed to be on the order of 10 km and do not affect the results reported here. Recent measurements by Pyle, however, suggest that the location errors may be significantly less than 10 km.
The contour plots in this paper are based on a grid of 3000 points, 60 along the x axis and 50 along the y axis. This is the only grid scale currently available in the analysis software. On maps of the United States, this grid corresponds to a spatial resolution of 90 km in the east–west direction and to 65 km in the north–south direction. Variations of the flash parameters occurring on a smaller scale will, of course, not be resolved in the present maps.
All results are summarized into two categories: ground flash counts and the percentage of cloud-to-ground flashes lowering positive charge to ground.
a. Ground flash counts
The results of the ground flash count for the years 1992–95 are presented in Table 1. Total flashes refer to the combined number of negative and positive flashes (columns 2–5) and positive flashes refer to just those cloud-to-ground flashes lowering positive charge to ground (columns 6–9). A plot of the total monthly flash values is presented in Fig. 2 and a plot for positive flashes in Fig. 4. Annual contour plots of the total cloud-to-ground lightning flashes for each year, 1992–95, are given in Figs. 3a–d. Similarly, annual positive cloud-to-ground lightning flashes are contoured in Figs. 5a–d for each year.
1) Total cloud-to-ground (positive and negative) flashes
Table 1, columns 2–5, reveals that the annual number of cloud-to-ground flashes in the contiguous United States increased from 16.2 million flashes in 1992 to 24.2 million flashes in each of the two subsequent years. This is approximately a 50% increase. In 1995, the total count decreased slightly to 22.3 million flashes. A plot of the monthly values in Fig. 2 shows that the peak month for total lightning flashes in each of the four years was July. The large increase in the total lightning for 1993 and 1994 is the result of the increased lightning counts in the summer months of June, July, and August. Variations in the spring and fall lightning counts have little effect on the total count, even though the percentage change can be large. For example, compare October 1992 to October 1993.
Note that the area of peak lightning flash density in the four years (Fig. 3) varies from Florida in 1992 to an area centered on St. Louis, Missouri, in 1993, and then back to Florida in 1994. The location of peak cloud-to-ground lightning flash density in 1993 (Fig. 3b) in the Midwest contrasts with the area peak flash density in Florida in each of the earlier years for which lightning flash density measurements have been available (Orville 1991, 1994). In 1995, the peak lightning cloud-to-ground flash densities were in southern Louisiana and in the area of the Kentucky–Illinois border.
2) Positive flashes
The number of cloud-to-ground positive flashes by month is listed in Table 1, columns 6–9, and plotted by month in Fig. 4 for 1992–95. The peak number of positive flashes occurred in July for three years (1992–94) and in May for 1995. The geographical distribution of the positive flashes is shown in the contour plots of Figs. 5a–d.
Note that the geographical distribution of the positive flashes is in the Midwest in all four years. A minimum of positive flashes is observed along the East Coast from the Carolinas north. A peak in the positive flashes is dramatically evident in northeastern Iowa in 1993. Clearly the geographic distribution of positive flashes is consistently different from the distribution of total flashes for all four years.
b. Percent positive polarity
A consideration of the number of negative and positive flashes leads us to examine the percentage of flashes that lower positive charge. Table 2 lists the percentage as a function of month and year. The annual percentage is shown to be on the order of 4%–5% (1992–94), but 9.3% in 1995. The monthly variation is greater. Figure 6 shows that the percentage of positive flashes ranges from 3% to 4% in August to more than 20% in December in 1992 and 1993.
There is a wide geographical variation of percentage positive. Figures 7a–d show a consistent pattern of percentage positive polarity that repeats itself in each of the three years (1992–94), but then increases overall in 1995. At lower latitudes, in the years 1992–94, the southern states have a percentage positive polarity that is 5% or less except for a relative maximum (of order 10%) in eastern Texas. In 1995, the percentage increases in some southern states to over 12.5%, but remains less than 25% on this grid scale. The highest percentage of positive cloud-to-ground lightning flashes occurs along the West Coast, followed by maxima in the upper Midwest and Maine.
a. Ground flash density
The dramatic variation of the ground flash density from maxima in Florida in 1989, 1990, 1991, (Orville 1994, plates 1a, 2a, and 3a), 1992 and 1994 (Figs. 3a and 3c) to a singular maximum in 1993 (Fig. 3b) in the Midwest is further evidence of the unusual weather that dominated the Midwest in 1993. In the words of Changnon (1994),
The flood of 1993 was a massive summer-long culmination of heavy rains encompassing the Upper Mississippi River Basin and the lower Missouri River Basin and their tributaries. Record summer rainfall over 9 midwestern states created record high river levels leading to astronomical economic losses and social disruption. The record floods were unique meteorologically because of their long duration through summer. The $15 to $20 billion in losses made the flood of 1993 the worst on record in the United States.
The 1993 ground flash density peak in the Midwest, exceeding 11 flashes per square kilometer, is consistent with the observations of the heavy rains throughout the summer of 1993.
Lightning and precipitation studies were initiated approximately three decades ago. Battan (1965) deployed approximately thirty rain gauges in the Santa Catalina Mountains of Arizona from 1957 to 1962 and discovered a statistically significant correlation between mean rainfall and the number of cloud-to-ground lightning flashes. Subsequent studies have been made by Piepgrass et al. (1982) and Shih (1988), both at the NASA Kennedy Space Center. Shih examined cloud-to-ground lightning flashes and precipitation observations and found that the flash rate predicted the total area-average rainfall. Kane (1990) performed an analysis, based on cloud-to-ground lightning data, of a localized flooding event that occurred in eastern Ohio on 14 June 1990. He noted a well-behaved correlation between the areas that received over 3" of rain and the most dense region of cloud-to-ground lightning flashes.
We should be careful, however, to note that the referenced studies and the data presented in this paper are for thunderstorms over land. Extrapolation of any results from land to thunderstorms over the ocean is not justified. Zipser (1994) has noted that in parts of West Africa, Oceania, and other equatorial regions, climatologically, there is actually a minimum of thunderstorm activity corresponding to the month(s) of peak precipitation. Clearly, more research is needed to understand the relationship between precipitation and lightning.
b. Positive ground flash density
The positive flash densities have maxima in the Midwest in all four years, 1992–95. It is particularly dramatic in 1993 where there is one maximum centered in eastern Iowa. The maximum of 0.9 flashes per square kilometer in Fig. 4b is three times the maximum in Florida, on the order of 0.3 flashes per square kilometer. Positive flash densities for the years 1992–95 are low along the East Coast and in the West, continuing this observation from the years, 1989–91 (Orville 1994).
It is interesting that the location of maximum positive flash densities, in the Midwest, is geographically in agreement with the reported locations of discharges between the top of clouds and the ionosphere, now known as sprites (Sentman and Westcott 1993; Winckler et al. 1993; Lyons 1994). This may be fortuitous. Searches for sprites in other geographical areas of the United States and other countries are continuing.
c. Percent positive polarity
The percentage positive lightning contours have local maxima along the West Coast, in the upper Midwest, and in the Northeast (Fig. 7). Instrumentation effects that might contribute to the local maxima along the West Coast and in the Northeast have been discussed previously by Orville (1994). No instrumentation effects, however, are known to contribute to the high percentage of positive lightning in the upper Midwest and in the southern Plains and lower Midwest in 1995 (Fig. 7d). The explanation for these secondary percentage positive maxima remains unknown.
We have noted previously (Orville 1994) that the broad area of positive polarity in the Midwest coincides with the observation by Rutledge and MacGorman (1988) of the occurrence of positive flashes in stratiform precipitation in mesoscale convective systems (MCSs), a frequent storm type east of the Rocky Mountains. The existence of thunderstorms east of the continental divide, dominated by positive lightning, is now well documented, for example, Curran and Rust (1992), Branick and Doswell (1992), and Stolzenburg (1994).
Cloud-to-ground lightning data for the years 1992–95 have been analyzed for the geographical distribution of flash density for total flashes, positive flashes, and the percentage of flashes that lower positive charge to ground. In summary, the results are as follows:
The total measured flash counts were 16.3 million (1992), 24.2 million (in both 1993 and 1994), and 22.3 million (1995).
The areas of maximum flash density were in Florida in 1992 (9–11 flashes per square kilometer) and 1994 (11–13 flashes per square kilometer), and in the Midwest in 1993 (11–13 flashes per square kilometer), coinciding with the floods that dominated the summer of 1993 in the Midwest. In 1995 the maximum was 9–11 flashes per square kilometer and it occurred along the Kentucky–Illinois border and in southern Louisiana.
Positive flash densities have maxima in the Midwest in all four years with values of 0.4 (1992), 1.0 (1993), 0.7 (1994), and 1.8 flashes per square kilometer (1995).
The annual mean percentage of lightning flashes that lowered positive charge was 4.2% (1992), 4.6% (1993), 4.9% (1994), and 9.3% (1995). The apparent sudden increase in the 1995 value clearly deserves further study and will be compared, with interest, to the 1996 value.
The monthly percentage of positive lightning flashes varies throughout the year, ranging from 3% (August 1992) to 25% (December 1993).
The percentage of positive lightning, averaged for the year, varies geographically, with maxima over 25% along the West Coast and Northeast, and over 12% in the upper Midwest and lower Midwest (1995).
All the cloud-to-ground lightning data from 1989 through 1995 are now being studied on a monthly basis to determine the lightning characteristics for an “average year” in the contiguous United States. This “average year,” in turn, will be composed of the 12 average months with the characteristic properties of cloud-to-ground lightning to include flash density, percentage positive polarity, multiplicity, and peak current.
The lightning data were obtained from the GeoMet Data Services, Incorporated, Tucson, Arizona. The interest and assistance of Ken Cummins in this project are greatly appreciated. We thank Barbara Orville for her editorial assistance. Data handling at Texas A&M University is under the direction of Jerry Guynes and Robert White and we thank them for their assistance. This research is part of a lightning program supported by the National Science Foundation (ATM-9213787) and the National Oceanic and Atmospheric Administration (Contract NA37WA0543).
Corresponding author address: Richard E. Orville, CIAMS, Dept. of Meteorology, Texas AM University, College Station, TX 77843-3150.