\PERGAMON Journal of Atmospheric and Solar!Terrestrial Physics 59 "0887# 864Ð868

S0253Ð5715:87 ,*see front matter Þ 0887 Elsevier Science Ltd[ All rights reservedPII] S 0 2 5 3 Ð 5 7 1 5 " 8 7 # 9 9 9 9 6 Ð 7

Comparison of preliminary breakdown pulses observed inSweden and in Sri Lanka

Chandima Gomes\ Vernon Cooray�\ Chandana JayaratneInstitute of High Voltage Research\ Uppsala University\ Sweden

Received 02 February 0886^ in revised form 02 January 0887^ accepted 1 February 0887

Abstract

In this paper we have analysed breakdown pulse trains preceding the _rst return stroke in 36 negative cloud!to!ground"CG# lightning ~ashes observed in Sri Lanka "tropics# and that in 30 negative CG ~ashes recorded in Sweden "temperateregions#[ In the data obtained in Sri Lanka\ breakdown pulses could be detected only in nine ~ashes[ In other ~asheseither these pulses are absent or\ if they are present\ their amplitudes should be below the background noise level[ Fromthe nine ~ashes with breakdown pulse trains\ we obtained the following results[ The ratio between the maximumbreakdown pulse amplitude and the return stroke amplitude "BP:RS ratio#\ as a percentage\ is 05[4)[ The time durationbetween the most active part of the pulse train and the return stroke "BP:RS separation#\ is 00[8 ms and the ratiobetween the noise amplitude and the return stroke amplitude "N:RS ratio#\ as a percentage\ for all 36 ~ashes\ is 4[9)[In the ~ashes observed in Sweden\ breakdown pulses are detectable in all records[ In this data\ the BP:RS ratio\ as apercentage\ is 090[3)\ the BP:RS separation is 02[7 ms and the N:RS ratio\ as a percentage\ is 2[6)[ All the abovevalues are arithmetic means[ Most of the pulses of breakdown pulse trains observed in both countries are bipolar innature with the initial polarity the same as that of the succeeding return stroke[ Þ 0887 Elsevier Science Ltd[ All rightsreserved[

0[ Introduction

The cloud to ground discharge is made up of oneor more intermittent partial discharges which involves acombination of leader and return strokes[ During thisperiod between return strokes*before the _rst returnstroke and after the last subsequent return stroke*mic!rosecond scale fast _eld variations\ as well as millisecondscale slow _eld changes\ can be seen in lightning gen!erated _eld traces[ In this paper we consider only thetrains of microsecond scale electric _eld pulses\ known asbreakdown pulses\ which precede the _rst return stroke[

There are several observations made on the pre!_rstreturn stroke pulse trains in temperate regions "Clarenceand Malan\ 0846^ Beasley et al[\ 0871^ Brook 0881^Ogawa\ 0882^ Gomes et al[\ 0886\ etc[#[ In some obser!vations\ these pulse trains were found with several pulses

� Corresponding author[ Tel[] 9935 07 422525^ fax] 993507491508^ e!mail] vernon[coorayÝhvi[uu[se

having amplitudes that are comparable to the amplitudeof the succeeding return stroke "Brook\ 0881#[ However\Brook has pointed out that there is a signi_cant di}erencein the amplitudes of pulse trains observed in Floridaand that in Albany\ New York[ Thus it is important toinvestigate the relationship between the amplitudes ofreturn strokes and that of corresponding preliminarybreakdown pulses\ in di}erent geographical regions[ Onemay also make a hypothesis\ from these characteristics\about the lightning initiating process\ inside the cloud[For example\ recently Cooray and Scuka "0885#\ sug!gested that the necessary condition to create a cloud!to!ground ~ash is the generation of a vertical channel belowthe negative charge centre[ The probability of such anevent increases if the breakdown possibility is highbetween the negative charge centre and the positivecharge pocket located at the cloud base[ Therefore\meteorological conditions which favour the productionof strong positive charge pockets at cloud bases enhancethe percentage of lightning ~ashes that terminate at theground[ Further they suggested that the strength of thepreliminary breakdown pulses can be used as a measure

C[ Gomes et al[:Journal of Atmospheric and Solar!Terrestrial Physics 59 "0887# 864Ð868865

of the strength of the positive charge pocket[ By analysinga small sample of data from Sri Lanka\ which was avail!able to them\ they concluded that preliminary breakdownpulse trains are much weaker in Sri Lanka\ compared tothose in Sweden[ Based on this study\ in turn\ they cameto the conclusion that the positive charge pocket is muchweaker in tropical clouds than it is in clouds in temperateregions\ and that may be the reason for the low per!centage of ground ~ashes observed in Sri Lanka[ How!ever\ we felt that the amount of data available to Coorayand Scuka "0885# was statistically not su.cient\ and ananalysis of a larger sample of data was needed\ beforemaking a conclusion regarding the characteristics of pre!liminary breakdown pulse activity in Sri Lanka and inSweden[

1[ Experimental procedure

Measurements were conducted in Uppsala "latitude48[7>N\ longitude 06[5>E#\ Sweden\ in July 0882 and inColombo "latitude 5[8>N\ longitude 68[75>E#\ Sri Lanka\in March:April 0885[ The negative cloud!to!ground "CG#~ashes obtained in Sweden pertained to several frontalthunderstorms on 11 July 0882[ These thunderstormsoccurred at a distance of 09Ð099 km from the measuringstation[ The negative CG ~ashes recorded in Sri Lankacorrespond to electric _elds associated with two con!vective thunderstorms\ on 15 March and 1 April 0885[The approximate range of distance to the places of light!ning strike from the measuring station was within 19 kmfor most of the strokes\ as thunder was heard in thesecases[ The cloud top heights during the thunderstormswere not known but\ according to meteorological data inSri Lanka\ the average height of the cloud tops duringthunderstorms in April is about 09Ð01 km[

The antenna system used at both measuring sites wasidentical to that described previously by Cooray andLundquist "0871#[ The vertical _eld was sensed by a ~atplate antenna[ After passing through a bu}er ampli_er\the signal was directly fed to a transient recorder\ througha properly terminated coaxial cable[ The zero!to!peakrise time of the output of the antenna system for a stepinput voltage was 9[1 ms[ The decay time constant in thesystem in Sweden was about 19 ms and that in Sri Lankawas about 4 ms[ Recording systems at both sites wereidentical to that described previously by Cooray andPerez "0883#[ This system consists of a LeCroy transientrecorded with 0 Megabyte memory[ The sampling timewas kept at 9[4 ms\ and the pre!trigger delay time was 099ms at the site in Sweden and 299 ms at the site in SriLanka[ Several of the _eld traces recorded in Swedenwere triggered by the preliminary breakdown pulse trainwhile the rest were triggered by the return stroke[ In SriLanka\ all the _eld traces recorded were triggered by thereturn stroke[

Fig[ 0[ "a# Flash no[ 82119613[SWE^ a part of the BP trainrecorded in Sweden^ "b# Flash no[ 85159293[SRI^ a part of thepulse train recorded in Sri Lanka[ In both "a# and "b#\ thehorizontal scale is in 49 ms divisions and the vertical scale is inthe same arbitrary units[

2[ Results and discussion

Figure 0"a# and "b# shows several pulses of a break!down pulse train recorded in Sweden and those of apulse train observed in Sri Lanka\ respectively[ The wavepro_le of breakdown pulses traced at both sites are simi!lar\ except that the second half\ compared to the initialhalf\ is larger in the pulse trains recorded in Sweden[ Theindividual pulse separation in the pulse train in Swedenis less than that of the pulse train observed in Sri Lanka[Almost all the pulses are bipolar in nature\ with initialpolarity as same as that of the succeeding return stroke[This may be evidence of the fact that the event whichgives rise to these pulses are the same at both locations[Two typical examples of a preliminary breakdown pulseburst and succeeding return stroke observed in Sweden

C[ Gomes et al[:Journal of Atmospheric and Solar!Terrestrial Physics 59 "0887# 864Ð868 866

are given in Fig[ 1 and those in Sri Lanka are given inFig[ 2[ One can clearly see that the amplitude of thepreliminary breakdown pulse train\ compared to that ofthe succeeding return stroke\ is much higher in Swedenthan that of pulse trains in Sri Lanka[ The ratio betweenthe amplitude of the breakdown pulse train and returnstroke amplitude "BP:RS ratio# of the ~ashes pertinentto Sweden are 68) "Fig[ 1"a## and 177) "Fig[ 1"b##\respectively[ The same ratio of the two ~ashes recordedin Sri Lanka are 08) "Fig[ 2"a## and 14) "Fig[ 2"b##\respectively[

In the data set obtained in Sweden\ all 30 ~ashes rec!orded contain well detectable breakdown pulse trainsprior to the _rst return stroke[ The arithmetic mean andthe geometric mean of the ratios between the maximumbreakdown pulse train amplitude and return stroke

Fig[ 1[ BP trains and succeeding return strokes recorded inSweden[ "a# Flash no[ 82119696[SWN^ the horizontal scale is in0 ms divisions^ "b# Flash no[ 82119615[SWE^ the horizontal scaleis in 4 ms divisions[ In both "a# and "b# the vertical scale is inarbitrary units[

Fig[ 2[ BP trains and succeeding return strokes recorded in SriLanka[ "a# Flash no[ 85919323[SRI^ the horizontal scale is in 1ms divisions[ "b# Flash no[ 85919308[SRI^ the horizontal scale isin 4 ms divisions[ In both "a# and "b# the vertical scale is inarbitrary units[

amplitude "BP:RS ratio#\ as a percentage\ are 090[3)and 37[4)\ respectively[ The maximum BP:RS ratio isas high as 516) and the minimum is 7[1)[ The abovetwo mean values of the noise amplitude to return strokeratios "N:RS ratio#\ as a percentage\ are 2[6) and 2[1)\respectively[ Therefore\ even the minimum BP:RS ratiois greater than twice the mean N:RS ratio[

In the data set recorded in Sri Lanka\ only nine outof 36 ~ashes consist of _rst return strokes preceded bybreakdown pulses with detectable level of amplitudes "i[e[only 08[1)#[ In the other ~ashes\ breakdown pulses areeither not present or\ even if they are present\ their ampli!tudes should be below the background noise level[ Inthese nine ~ashes with breakdown pulses visible\ thearithmetic mean and the geometric mean of the BP:RSratios are 05[4) and 03[5)\ respectively[ The maximum

C[ Gomes et al[:Journal of Atmospheric and Solar!Terrestrial Physics 59 "0887# 864Ð868867

Table 0

N:RS ratio as aBP:RS ratio as a percentage ")# percentage ")# BP!RS separation "ms#

No[ of ~ashesData set Total no[ of with BP Arith[ Geom[ Arith[ Geom[ Arith[ Geom[from ~ashes detectable mean mean Max Min mean mean mean mean Max Min

Sweden 30 30 090[3 37[4 516[0 7[1 2[6 2[1 02[7 7[6 69[9 1[9Sri Lanka 36 98 05[4 03[5 15[3 5[1 4[9 3[5 00[8 8[7 12[9 2[4

and the minimum of the above ratios are 15[3) and5[1)\ respectively[ The two mean values of the N:RSratios are 4[9) and 3[5) respectively[ Therefore\ it canbe noticed that the minimum BP:RS ratio is not muchhigher than the mean N:RS ratio[ So\ one can concludethat all the ~ashes\ other than the nine ~ashes we men!tioned earlier\ may consist of breakdown pulses withamplitudes less than that of the background noise[ Theresults are summarised in Table 0[

The frequency distribution of the BP:RS ratios of ~a!shes observed in Sweden is given in Fig[ 3"a#[ A largefraction of ~ashes contain breakdown pulse trains ofwhich the amplitudes are greater than that of the suc!ceeding return stroke\ although the highest frequencybelongs to the 09Ð19) range[ The same distribution ofthe data obtained in Sri Lanka "only nine ~ashes# is givenin Fig[ 3"b#[ The sample size of this data is not su.cientto extract statistical results[ However\ one can see thatthe maximum BP:RS ratio is 15[3)\ which is much lessthan the corresponding value for the Swedish data[

The time separation between the most active part of

Fig[ 3[ "a# Frequency distributions of BP:RS ratio for ~ashesrecorded in Sweden^ "b# Same distribution for ~ashes recordedin Sri Lanka[ Note that the sample size of the data set from SriLanka is considerably smaller than that from Sweden[

Fig[ 4[ "a# Frequency distribution of BP:RS separation of the~ashes recorded in Sweden^ "b# the same distribution of ~ashesrecorded in Sri Lanka[

the breakdown pulse train "usually the centre of the pulsetrain# and the succeeding return stroke "BP:RS separ!ation#\ in the case of the Swedish data\ has an arithmeticmean and a geometric mean of 02[7 ms and 7[6 ms\respectively[ In more than half of the ~ashes\ the BP:RSseparation is in the range of 9Ð09 ms "Fig[ 4"a##[ In thereading set obtained in Sri Lanka\ the BP:RS separationhas an arithmetic mean and a geometric mean of 00[8 msand 8[7 ms\ respectively\ while the maximum is 12 msand the minimum 2[4 ms[ The range with the highestnumber of ~ashes is 4Ð09 ms "Fig[ 4"b##[ These resultsshow that the BP:RS separations in both countries arenot very di}erent from each other "Table 0#\ although\as mentioned earlier\ the sample size is too small to givecomments on statistical facts pertinent to the breakdownpulses observed in Sri Lanka[

3[ Conclusions

It is shown that in 099) of the ~ashes recorded inSweden\ breakdown pulse trains\ with detectable ampli!

C[ Gomes et al[:Journal of Atmospheric and Solar!Terrestrial Physics 59 "0887# 864Ð868 868

tudes\ precede the _rst return stroke[ In Sri Lanka\ suchpulse trains were observed in less than 19) of the totalnumber of ~ashes recorded[ Even if we consider the~ashes with breakdown pulses detectable\ recorded in SriLanka\ the arithmetic mean of the BP:RS ratio is aboutsix times less than that for the data recorded in Sweden[Hence\ it is clearly understood that in the tropics theinitial breakdown process in lightning is much less pro!nounced than that in temperature regions[ One possiblereason for this is the comparatively weak positive chargepocket at the cloud base in the tropics[ This results maygive support to the suggestions made by Cooray andScuka "0885#[ However\ one has to take into account thefact that\ apart from being in the tropics\ Sri Lanka is asmall island in the Indian Ocean and our measuring sitewas also located very close to the coast[ Hence\ one hasto conduct further investigations to see whether theobservations which we have presented in this paper aredue to the tropical location of Sri Lanka or due to itsoceanic location[ Further studies should also be con!ducted to investigate the in~uence of the strength of thepositive charge pocket of thunderstorms on the ground~ash density of a region[

It should also be mentioned that the two data setsemployed in this study pertain to di}erent types of thun!derstorms^ in Sweden they are frontal thunderstorms andin Sri Lanka convective thunderstorms[ At present weare not in a position to discuss the in~uence of these twotypes of thunderstorm on our results[

As we had no facilities in Sri Lanka to _nd the locationof lightning strikes\ we are not in a position to give aquantitative estimation on the in~uence of propagatione}ects on our results[ However\ the following argument\which treats the matter qualitatively\ may justify the vali!dation of our conclusions[ In Sri Lanka\ where most ofthe lightning ~ashes struck within approximately 19 kmfrom the measuring site\ as thunder was audible[ InSweden\ the distance to the lightning varies from closerange to about 099 km[ In the Swedish data set\ all the~ashes within 19 km possess considerably intense pre!liminary breakdown pulse trains\ while the Sri Lanka

data set for the same range of distance\ has very few~ashes with preliminary breakdown pulses*and they arealso much lower in intensity[ Furthermore\ in this study\we analysed the relative amplitude of preliminary break!down pulse trains "with respect to the succeeding returnstroke# but not their absolute amplitudes[ Hence\although it is strictly not correct\ we may expect theBP:RS ratio to remain more or less the same over adistance of a few tens of kilometres[

Acknowledgements

Financial assistance by the IPPS of the InternationalScience Programs\ Uppsala University\ Sweden\NARESA for research grant number RG:84:P:91\ andSwedish Natural Science Centre for research grant num!ber G!AA:GU 90337!209 are gratefully acknowledged[

References

Beasley\ W[H[\ Uman\ M[A[\ Rustan\ P[L[\ 0871[ Electric _eldspreceding cloud to ground lightning ~ashes[ J[ Geophys[ Res[76\ 3772Ð3891[

Brook\ M[\ 0881[ Breakdown electric _elds in winter storms[Res[ Lett[ Atmos[ Electr[ 01\ 36Ð41[

Clarence\ N[D[\ Malan\ D[J[\ 0846[ Preliminary discharge pro!cesses in lightning ~ashes to ground[ Quart[ J[ Roy[ Met[ Soc[72\ 050Ð061[

Cooray\ V[\ Lundquist\ S[\ 0871[ On the characteristics of someradiation _elds from lightning and their possible origin inpositive ground ~ashes[ J[ Geophys[ Res[ 76\ 00\192Ð00\103[

Cooray\ V[\ Perez\ H[\ 0883[ Some features of lightning ~ashesobserved in Sweden[ J[ Geophys[ Res[ 88\ 09\530Ð09\541[

Cooray\ V[\ Scuka\ V[\ 0885[ What attracts a lightning toground< Proc[ Int[ Conf[ Atmos[ Elect[\ Osaka\ Japan\ 145Ð148[

Gomes\ C[\ Thottappillil\ R[\ Scuka\ V[\ 0996[ Bipolar electric_eld pulses in lightning ~ashes over Sweden[ Proc[ 01th Int[Zurich Symp[ EMC\ 20F3[

Ogawa\ T[\ 0882[ Initiation of lightning in clouds[ J[ Atmos[Elect[ 02\ 010Ð021[