~ Pergamon Journal of Atmospheric and Terrestrial Physics, Vol. 57, No. 4, pp. 325-331, 1995

Elsevier Science Ltd Printed in Great Britain

0021-9169(94)E0003--6 0021-9169/95 $9.50+0.00

Observations of lightning flash rates and rain-gushes in Gaborone, Botswana

E. R. JAYARATNE, V. RAMACHANDRAN and K. R. S. DEVAN

Department of Physics, University of Botswana, Gaborone, Botswana

(Received in final form 16 December 1993 ; accepted 21 January 1994)

Abstraet--A CGR3 lightning flash counter has been used to obtain data over three fightning seasons between 1990 and 1993 in Gaborone, Botswana. This counter has a range of about 14 km and is able to discriminate between intracloud flashes and negative and positive ground flashes. The register outputs were interfaced to a microcomputer so that occurrences of the various types of flashes could be recorded in real- time. The results showed that the ratio of intracloud flashes to ground flashes was a maximum in the early and later stages of a thunderstorm, being lowest during the more active intermediate periods. The total flash rate was determined by counting the number of flashes within intervals of 5 min during the course of several intense storms that occurred during the period of observation. It was found that the ratio of intracloud flashes to ground flashes decreased as the flash rate increased. Very active storms with high flash rates yielded a relatively larger proportion of ground flashes. Heavy precipitation was very often measured at the ground 2-6 min after overhead lightning, confirming the existence of the well-known rain-gush phenomenon. It is argued that this time interval is too short for the rain to be a consequence of the lightning flash, and it is suggested that the lightning originated on the falling precipitation in accordance with a precipitation-based mechanism of thunderstorm electrification.

l. INTRODUCTION

Lightning flashes are broadly classified into two types, ground flashes and intracloud flashes. Most ground flashes carry a net negative charge but positive ground flashes are not uncommon. The ratio of intracloud flashes to ground flashes, Z, has been experimentally determined at several locations around the world. PIERCE (1970) and PRENTICE and MACKERRAS (1977) reviewed the results of several previous studies and derived empirical formulae relating Z to geographical latitude. They showed that Z varied from about 6 to 9 at the equator to about 2 at temperate latitudes. This dependence was attributed to the height of the freezing level above the ground. Within a thun- dercloud, most of the charge centres, on which light- ning flashes originate, are located in the ice phase (LATHAM, 1981; WILLIAMS, 1989). PIERCE (1970) and PRENTICE and MACKERRAS (1977) suggested that, since the freezing level occurs at higher altitudes in the tropics, intracloud flashes occur more readily than ground flashes (Z high), while in the temperate zones, the proximity of the freezing level to the ground gives rise to a higher proportion of ground flashes (Z low). Recently, PRICE and RIND (1993) have proposed an alternative hypothesis for the variation of Z with lati- tude. They suggest that Z is principally determined by

the volume of the mixed-phase region of a cloud (from 0°C to cloud-top). It is well known that strong elec- trification occurs in such regions where graupel, ice crystals and supercooled droplets coexist above the melting level (LATHAM, 1981). Strong convective motions within tropical thunderclouds give rise to higher cloud-tops and larger mixed-phase volumes. The occurrence of intracloud lightning is enhanced due to dielectric breakdown in such regions, resulting in higher Z values (WILLIAMS et al., 1989). RUTLEDGE et aL (1992) studied four tropical thunderstorms in Darwin, Australia and found that Z increased with the total flash frequency, while the height of the freez- ing level remained approximately constant. They con- cluded that Z and the total flash frequency were both positively correlated with the volume of the mixed- phase region in the clouds.

CGR3 lightning flash counters are specifically designed to discriminate between intracloud flashes and ground flashes, and have been recently used at several sites between the latitudes of 60°N and 27°S. MACKERRAS and DARVENIZA (1992) surveyed the results and concluded that there was no consistent relationship between Z and latitude as had been thought before. They attributed the apparent latitude dependence in early studies to inaccurate methods of measurement. Moreover, thunderstorms at a given

325

326 E. R. JAYARATNE et al.

location often exhibit marked differences in Z. For example, MACKERRAS (1985) showed that Z for thunderstorms in Brisbane, Australia, varied from 0.9 to 24.7.

WILLIAMS et al. (1989) showed that intracloud activity generally dominates the early and later stages of a thunderstorm, with cloud to ground activity being predominant in-between.

It is known that intense falls of precipitation are often observed on the ground shortly after a nearby lightning flash to ground. This so called 'rain-gush' phenomenon has raised an important question as to whether the lightning caused the rain or vice-versa. Early workers thought that the charged particles were levitated in the strong thunderstorm electric fields until a lightning flash destroyed the field, allowing them to precipitate. This theory has long been discarded. MOORE et al. (1962, 1964) proposed that a negative ground flash deposits some positive charge in the cloud. This charge moves out encountering the negatively charged cloud particles to which it becomes attached, passing on a far stronger positive charge. The positively charged particles move through the cloud attracting the negatively charged cloud droplets and growing rapidly until they are large enough to fall. These large cloud particles make up the sudden gushes of rain and hail following lightning. JAYARATNE and SAUNDERS (1984) have suggested an alternative hypothesis where the lightning flash is caused by the falling precipitation. The theory is based on a series of laboratory experiments conducted by JAYARATNE et al. (1983) who showed that, during ice crystal col- lisions, a riming graupel pellet acquires a net negative charge at a typical cloud water content of 1 g m -3 when the temperature was colder than about - 2 0 ° C and a positive charge when it was warmer. JAYARATNE and SAUNDERS (1984) suggested that a sudden increase of the precipitation rate in the form of falling graupel from above the freezing level would accordingly give rise to a pocket of positive charge. Lower positive charge centres associated with falling precipitation have been observed before (SIMPSON and SCRASE, 1937; KUETTNER, 1950). CLARENCE and MALAN (1957) suggested that this positive charge centre may trigger cloud to ground lightning flashes from the larger negative charge centre above. JAYARATNE and SAUNDERS (1984) pointed out that the very same grau- pel particles that gave rise to the ground flash may fall as a rain-gush on the ground a few minutes later.

Note that in both theories, the rain-gush reaches the ground after the nearby lightning flash. However, the time difference between the two is crucial to our understanding of the mechanisms involved. In the Moore et al. theory, the charges have to get onto

the cloud particles which must then move some way through the cloud before they can grow and become large enough to reach the ground as rain. This may take several minutes. The Jayaratne-Saunders theory does not require a long time interval. The positively charged precipitation may have fallen even below the cloud base when the ground flash is triggered, allowing the rain-gush to be observed at the ground within a few minutes of the associated lightning flash.

The lightning season in Gaborone, Botswana, begins in September, reaches its peak between November and January, and lasts up to about April. Lightning is very rarely observed during the winter months. The present study was conducted over the three lightning seasons spanning 1990-1993. Twenty- two major storms were monitored in real time to deter- mine the times of flashes, flash rate and rainfall rate. The time intervals between overhead lightning and instances of peak rainfall rate were checked in order to verify the two rain-gush theories above.

2. METHODS

A CGR3 lightning flash counter, developed by the University of Queensland, Brisbane, Australia, has been used to acquire data over the last three lightning seasons 1990-1993. The antenna was constructed to be compatible with CIGRE standards. It consisted of 18 m of 3 mm diameter insulated wire wrapped in eight loops around a vertical PVC pipe of external diameter 50 mm and length 3.3 m fixed with its lower end 1.75 m above a flat open horizontal ground. The counter is able to discriminate between the various types of lightning flashes by a voltage pattern rec- ognition technique (MACI~ERRAS, 1985). The unit includes three registers for negative ground flashes (NGF), positive ground flashes (PGF) and intracloud flashes (CF), each with an effective horizontal detec- tion range of about 14 km. Very close flashes, within a range of about 2 km are, in addition, identified on a fourth register as overhead flashes (OF). Outputs from the register unit were interfaced to a micro- computer and all flashes were recorded in real time. The rainfall rate was continuously recorded by the microcomputer which monitored the tip-times of a tipping bucket rain gauge placed near the antenna.

3. RESULTS

3.1. Flash rate

During the period of observation, we were able to record complete data from 22 major storms. Each of these storms were then divided into intervals of 5 min.

(a )

( c )

10

LL O 1 II

N

0.1 1

10

1 9 9 0 / 9 1

10

F l a s h e s p e r 5 m in in te rva l

1 9 9 2 / 9 3

Lightning and rain-gushes

1 O0

( b )

0 II

u

10_

0.1

327

1 9 9 1 / 9 2

::::: .............. . . . . . . . . . . . . . . . . . . . ~ ............. . i i ' i ' l I . i ': i I i i l i

............. i i . ~.....; .; = ,= i ................... ; ............ .d. i i ~. i l • i i

[ E ,, i i i i ,, , " , , i i i :::::::::::::::::::::::::::::::::::~-:: :::::: ~::: :;:::::: ~ :: :i:::: ::::::::::::::::::::::i:::::::::, t::::::: ::::::+t: ::::::: ~::::::::i:::::: ~::::" .~::::~:7:~::

. . . . . . . . . . i ......... i f f ' : ] i i . . . . . . . .

. . . . . . . i ............... i t-" 7 t ' i - t ................... R ' [ ~ t ~ ! " t ! ~

.................... i ............. ; i i [ i .............. i ........ ~ i ~ i ! . . . . . . . . . . . . . i ............... i . - .................... i .......... i . . ~ . i. ~ 7 = . { .

10 100

Flashes per 5 ra in in te rva l

( d ) 10

1 9 9 0 / 9 3

Ii O II v

N

E "

LL O II

N

0 . t 0.1 1 10 100 1 10

F l a s h e s p e r 5 m in in te rva l F l ashes p e r 5 m in in te rva l

Fig. 1. Z vs. flash rate in all storms for which data was recorded in (a) 1990-1991, (b) 1991-1992 and (c) 1992-1993. (d) Shows data from all 22 storms recorded over the three seasons.

1 O0

The number of ground flashes (GF = NGF+ PGF), intracloud flashe,~ (CF) and total flashes (TF = GF+ CF) was determined in each 5 min inter- val, and TF was taken to be a measure of the total flash rate. Figure 1 (a) shows the ratio of intracloud flashes to ground flashes (Z = CF/GF) for all 5 min periods as a function of flash rate (total flashes in that period) for the storms we were able to record in full during the season 1990-1991. Figures l(b) and (c) show the data for 1991-1992 and 1992-1993, respec- tively, while Fig. 1 (d) shows the data for all 22 storms recorded over the three seasons. In these four graphs, each value of Z plotted has been calculated from all flashes recorded in all 5 min intervals having the same number of total flashes. Thus, the value shown is the mean value of Z for each flash rate. I f there is no point shown on the graph, it implies that there were no 5 min intervals with that particular flash rate. Five- minute intervals with no GFs and/or CFs were ignored in the analysis. Each storm was demarcated by a 5 min interval with TF = 0 at either end.

3.2. Flash type

The durat ion of 1:he storms recorded varied from about 25 min to 2 h. In order to compare the temporal

distributions of the various types of flashes during the course of the storms, it was decided to normalise each storm to l0 units of time. For each storm, Z was calculated for each of the time units. Figure 2(a) shows the mean value of Z for each of the 10 time units. Thus, for example, the mean value o f Z in the first 10% of duration of the 22 storms studied was about 2.7. It should be pointed out, however, that the actual values of Z in these storms had a wide variation from about 0.2 to 8.0. Similarly, Fig. 2(b) shows the pro- port ion of the total number of flashes, TF, for each time unit averaged over all the storms. Thus, for exam- ple, it is seen that, on the average, approximately 5.5% of the flashes in a storm occurred during the first 10% of its total duration, and so on. A similar pattern was observed in most individual storms. Figure 3 shows Z, TF and GF as a function o f time, calculated to a resolution of 5 min, during a typical storm that occurred on 9 November 1992 between 02.50 h and 03.25 h.

3.3. Rain-gushes

In many of the close storms observed, sudden gushes of rain were detected at the ground soon after overhead ground flashes. Figure 4 shows three such

328

tJ_ o II

2 N

l

Time 18

12

U_

I--

0

Time

Fig. 2. Mean temporal variation of (a) Z and (b) the per- centage of total flashes, TF, during the 22 storms studied. Each storm has been normalised to 10 time units on the

horizontal axis.

examples that occurred on (a) 9 November 1992, (b) 14 March 1991 and (c) 16 February 1992. The histo- grams show the rainfall rate on the ground com- puted from the tip-times of the rain gauge. The arrows show the times at which the overhead flash register (OF) identified a negative ground flash as being within an effective horizontal range of about 2 km. The time interval between an overhead flash and the subsequent peak rainfall rate at the ground is crucial to our under- standing of the charging phenomenon. As described in the introduction, the Moore et el. theory requires a longer time interval than the Jayaratne-Saunders theory. This study enabled us to determine the rel- evant times in order to verify the plausibilities of the two hypotheses.

4. ANALYSIS AND DISCUSSION

Figure 1 shows very clearly that Z decreases with TF. The pattern is obvious in all four graphs shown. A similar trend is seen in the temporal distribution of flashes during storms. In agreement with WILLIAMS et

E. R. JAYARATNE et al.

al. (1989), Fig. 2(a) shows that Z is a minimum during the middle of a storm. This is also seen in many indi- vidual storms such as that shown in Fig. 3. As expected, the total flash rate is a maximum during the middle of a storm (Figs 2(b) and 3). Taken together, these two observations confirm our assertion that Z decreases as the total flash rate increases.

RUTLEDGE et al. (1992), on the basis of data from four individual storms, showed that Z increased with total flash rate. Higher flash rates were generally a consequence of increased intracloud activity in the active periods of the storms. Contrary to this, in the present study, the higher flash rates almost always corresponded to increased cloud to ground activity, especially during the middle of the storms. Many of the very intense storms gave low values of Z and were characterised by unusually large numbers of ground flashes. This is further illustrated in Fig. 5, which shows the counter data for all storms with TF > 10 that occurred during the three seasons. Note that Z decreases as the number of flashes in a storm increases. The two most intense storms that occurred (TF > 1000) consisted of over two ground flashes for every intracloud flash. Many of the storms with high Z exhibited low flash rates.

Figure 4 shows observations of rain-gushes at the ground shortly after overhead lightning. In Fig. 4(a), heavy rain began to fall within 2 min of the first over- head negative ground flash. This was followed by four more overhead flashes and a peak rainfall rate of 42 mm h- ~. In Fig. 4(b), only one overhead flash was recorded, and the rainfall rate showed a sharp peak of 100 mm h-~ just 2 min later. In Fig. 4(c), overhead flashes occurred at 23 : 53, 23 : 55 and 23 : 58 h with

:10

O2c Z N

10

o so 6 1'o 2'o Time (rain after the hour)

Fig. 3. Data from a storm that occurred on 9 November 1992 between 02.45 and 03.30 h, showing the temporal variation of Z, TF and GF. All points shown are averages over 5 min

intervals.

Lightning and rain-gushes 329

50-

(a) 9 Nov 92

E E

r r =

¢- " 3

I T

40-

30-

20

10

10

120.

12 14 16 r ime (min past the hour)

(b) 14 M a r 91

18

=_. E 80

G)

r r

"E 4O

t r

6 8 10 12 14 Time (min past the hour)

(C) 16 F e b 92 5 0 - ~

t -

30- Q)

~ .

¢'r" 20-

E •

52 54 56 58 0 2 Time (min past the hour)

Fig. 4. Rainfall rates during storms observed on (a) 9 Nov- ember 1992, 03.084)3.2,0 h, (b) 14 March 1991, 19.00-19.15 h and (c) 16 February 1992, 23.50~0.04 h. The arrows indicate the times of occurrence: of overhead negative ground flashes.

the first rain fallin~; at the ground at 23:59 h and peaking shortly after 00:00 h. In all these obser- vations, the rain-gush reached the ground between 2 and 6 min of an overhead negative ground flash. The height of the freezing level in Gaborone is about 5 km above the ground. To fall this distance in 2-6 min

12

. . . . i i l i i i J • i i i i ~ ~ i l • = ,

................ i i - i " " ........ [ ................... [ - i - T [ .......................... - ' T i ..... i i

. ~ i . , , = i F ' i o , . !! N 4 . . . . . ~ --lll" l ~ l l l ll ~ -- ~ j l l l l ~ ....... ~ . . . . . . l l ' ~ ...............

i C i • ! i u ! I F • • l " i , , , i,,E i i

" E ! = = H i . !

0 , . r , , ,=,, , , , , , , ,~, , , i , ! , i ,~ 10 100 1000 10OOO

TF Fig. 5. Zvs. TFfor all storms with more than 10 flashes that occurred during the three seasons. Each point corresponds to a single storm. The solid line represents Z = 1. These data are from the counter registers and not the microcomputer.

requires terminal fall velocities of 42-14 m s -t , respec- tively, values much greater than those of typical millimetre-sized heavy precipitation particles. For example, raindrops of diameter 1 and 2 mm have ter- minal velocities of about 3 and 13 m s -~ in the sub- cloud environment. Owing to their lower densities, graupel pellets have even lower fall velocities. Con- sidering the short times observed between overhead lightning and the rain-gush, it seems reasonable to believe that in some cases, the precipitation had fallen past the freezing level when the lightning flashes occurred. The lower values of the measured time inter- vals between the lightning and the rain-gush are far too short for the Moore et al. mechanism to operate. According to this theory, a negative ground flash dumps some positive charge in the cloud. Negative ground flashes originate at temperatures between about - 10°C and - 20° C (LATHAM, 1981 ; KREHBIEL et aL, 1979 ; WILLIAMS, 1989). These levels correspond to heights of 6-7 km in Gaborone during the summer. Assuming that the positive charge is deposited close to the negative charge centre, the minimum measured time interval between the lightning flash and the rain- gush is too short for the precipitation to fall to the ground. Moreover, this does not take into account the time required for the enhanced coalescence process to occur within the cloud before the precipitation may start to fall. Further evidence against the Moore et aL theory is presented by MARSH and MARSHALL (1993), who report substantial quantities of positively charged precipitation 9articles comprising a lower positive charge centre In a thundercloud before the first light- ning flash.

The short time lag observations may, however, be explained in terms of the Jayaratne-Saunders theory

330 E. R. JAYARATNE et al.

where the lightning flash is initiated by the falling precipitation. According to the laboratory results of JAYARATNE et al. (1983), graupel acquires positive charge during interactions with vapour-grown ice crystals between the temperature levels of 0°C and -20°C. The accumulated net positive charge on a falling graupel attains its maximum value at the freez- ing level. A burst of falling graupel would momen- tarily give rise to a lower positive charge centre that may initiate one or more negative ground flashes from the main charge centre above as described in the intro- duction. The positive charge carried by the falling precipitation may last well after falling past the freez- ing level. Thus, the ground flash may occur with the rain-gush quite close to the ground so that the time difference between the flash and the rain could be relatively short.

In order to verify the reliability of the instrument, some studies were carried out with a second CGR3 counter at the same site. Observations were also made with a horizontal wire antenna at another site about 1 km away. The various combinations of counters and antenna gave very consistent results. Visual obser- vations of lightning confirmed that the counters iden- tified intracloud and ground flashes correctly with an accuracy rate of about 90%. For example, on average, of every 10 flashes that we were able to visually con- firm as being either a ground flash or an intracloud flash, nine were identified correctly by the counter. All the results presented in this paper have been obtained with the standard antenna described in the intro- duction. Thus, the data may be compared with those from studies at other locations using the same type of counter and antenna.

5. CONCLUSIONS

The ratio of intracloud flashes to ground flashes decreases with total flash rate. The more active storms in Gaborone invariably consist of a larger proportion of ground flashes. During the course of individual storms, the ratio is generally highest in the early and later stages, falling to its lowest values during the more active intermediate stages.

Precipitation rates at the ground commonly peak 2-6 rain after an overhead negative ground flash, con- firming the existence of the well known 'rain-gush' phenomenon. Consideration of the terminal fall vel- ocities indicates that the falling precipitation is often below the freezing level when the flash occurs. This suggests that the precipitation is not a consequence of the lightning flash as proposed by MOORE et al. (1962, 1964). The Jayaratne and Saunders (1984) hypothesis, on the other hand, assumes that the flash is initiated by the positively charged graupel pellets comprising the commonly observed lower positive charge centre. The enhanced local electric field around this positive charge pocket may trigger a ground flash from the main negative charge centre above. At this instant, the falling precipitation may be relatively close to the ground and could result in a rain-gush within the shortest time intervals observed in the present study.

Acknowledgements--This work was supported by a grant from the University of Botswana Research and Publications Committee. We are grateful to Dr Dave Mackerras of the University of Queensland for his advice and support and the Botswana Meteorological Services Department for allowing us to install and operate our equipment on their premises.

CLARENCE N. D. and MALAN D. J.

JAYARATNE E. R. and SAUNDERS C. P. R.

JAYARATNE E. R., SAUNDERS C. P. R. and HALLETT J.

KREHBIEL P. R., BROOK M. and McCRORY R. A.

KUETTNER J.

LATHAM J.

MACKERRAS D.

MACKERRAS D. and DARVENIZA M.

REFERENCES

1957 Preliminary discharge processes in lightning flashes to ground. Q. J. R. Meteorol. Soc. 83, 161-172.

1984 The 'rain-gush', lightning and the lower positive charge centre in thunderstorms. J. geophys. Res. 89, 11816-- 11818.

1983 Laboratory studies of the charging of soft-hail during ice crystal interactions. Q. J. R. Meteorol. Soc. 109, 609-630.

1979 An analysis of the charge structure of lightning dis- charges to ground. J. geophys. Res. 84, 2432-2456.

1950 The electrical and meteorological conditions inside thunderclouds. J. Meteorol. 7, 322-332.

1981 The electrification of thunderstorms. Q. J. R. Meteorol. Soc. 107, 277-298.

1985 Automatic short-range measurement of the cloud flash to ground flash ratio in thunderstorms. J. geophys. Res. 911, 6195-6201.

1992 Progress report on world-wide survey of cloud-flash to ground-flash ratio using CGR3 instruments.

MARSH S. J. and MARSHALL T. C.

MOORE C. B., VONNEGUT B., MACHADO J. A. and SURVILAS H. J.

MOORE C. B., VOtqNEGUT B., VRABLIK E. A. and MCCRAIG D. A.

PIERCE E. T.

PRENTICE S. A. and MACKERRAS D.

PRICE C. and RIND D

RUTLEDGE S. A., WILLIAMS E. R. and KEENAN T. D.

SIMPSON G. C. and SCRASE F. J.

WILLIAMS E. R.

WILLIAMS E. R., WEBER M. E. and ORVILLE R. E.

Lightning and rain-gushes 331

Proc. Ninth International Conf. Atmos. Electricity. St Petersburg, Russia. pp. 380-383.

1993 Charged precipitation measurements before the first lightning flash in a thunderstorm. J. 9eophys. Res. 98, 16605-16611.

1962 Radar observations of rain-gushes following overhead lightning strokes. J. geophys. Res. 67, 207-220.

1964 Gushes of rain and hail after lightning. J. atrnos. Sci. 21,646-665.

1970 Latitudinal variation of lightning parameters. J. app. Meteorol. 9, 194-195.

1977 The ratio of cloud to cloud-ground lightning flashes in thunderstorms. J. app. Meteorol. 16, 545-550.

1993 What determines the fraction of cloud-to-ground light- ning in thunderstorms? Geophys. Res. Lett. 20, 463- 466.

1992 The down under Doppler and electricity experiment (DUNDEE): Overview and preliminary results. Bull. Amer. Meteorol. Soc. 73, 3-16.

1937 The distribution of electricity in thunderclouds. Proc. R. Soc. A161, 309-352.

1989 The tripole structure of thunderstorms. J. 9eophys. Res. 94, 13151 13167.

1989 The relationship between lightning type and convective state of thunderclouds. J. geophys. Res. 94, 13213- 13220.