JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 90, NO. A10, PAGES 9815-9823, OCTOBER 1, 1985 Electrical Measurements in the Atmosphere and the Ionosphere Over an Active Thunderstorm 1. Campaign Overview and Initial Ionospheric Results M. C. KELLEY, • C. L. SIEFRING, • R. F. PF•, x P.M. KiNTNER, • M. LARSEN, • R. GREEN, • R. H. HOLZWORTH, 2 L. C. HALE,3 J. D. MITCHELL, 3 ANDD. LE VINE'* The first simultaneouselectricfield observationsperformedin the ionosphereand atmosphereover an active nighttime thunderstormare reported here. In the stratosphere,typical storm-relateddc electric fieldswere detectedfrom a horizontal distanceof ~ 100 km, and transientelectricfields due to lightning were measuredat severaldifferent altitudes.In the ionosphereand mesosphere,lightning-inducedtransientelectricfieldsin the range of tens of millivolts per meter were detectedwith rise times at least as fast as 0.2 ms and typical duration of 10-20 ms. The transientshad significantcomponentsparallel to the magnetic field at 150 km altitude. This implies that either considerable Joule heating occurs or a collectiveinstabilityis presentbecauseof the high drift velocitiesinducedby the transientelectricfields. Copious numbersof whistlerswere genratedby the storm and were detectedabove but not below the baseof the ionosphere. We presentherethe outlineof a newmodelfor directwhistlerwavegeneration over an active thunderstormbased on these observations.The intensity of the observedtwo-hop whistlers implies that they were amplified along their propagation path and suggeststhat particles were precipitatedin both hemispheres. kHz. The electricfield detectorsconsistedof six sphericalsensors(1 foot in diameter) mounted on three setsof orthogonal Lightning, and thunderstormsin general, involve a variety booms (length 3 m) [Mozer and Serlin, 1969]. Conductivity of fascinating physical processes.The experimental program measurementsusingthe relaxation time constantmethod were describedhere was designedto determinethe effectsof this obtainedevery 30 min. Similar instrumentationpackageshave meteorologicalphenomenonon the ionosphereand also to been flown beforeand are describedby Holzworth [1981'! and investigate electrical effects in the stratosphere and meso- Holzworth and Chiu [1982]. More details on the ballooon sphere. To this end, a thunderstorm electric field campaign systemare given in the companionpaper [Holzworth et al., was organized and carried out during an active nighttime this issue],hereinafterreferred to as paper 2. thunderstorm which moved through the rocket range at the The Super-Arcusparachutepaylod carried a blunt-probe dc NASA Wallops Flight Center, Wallops Island, Virginia (latielectricfield sensor(with a maximum frequencyof about 100 tude = 37.8ø,longitude = 284.5ø). Hz) and a Gerdien condenserto measureatmosphericconducA 10-cm radar located at Wallops Island was used to moni- tivities. More detailed information on this measurement contor the backscatterintensitydue to hydrometeors.The radar, figurationcan be found in the works by Hale et al. [1981'! and along with visual and RF observations,yielded a good indica- Mitchell et al. [1982] and in paper 2. tion of thunderstormactivity. A ground-basedflat plate anBoth ballisticpayloadswere instrumentedwith dipole electenna also provided valuable diagnosticdata. tric field detectorsemploying sphericalsensorsmounted on Two high flying ballistictrajectorypayloadswere developed the end of extendablebooms. The ionosphericrocket, known to make measurements in the mesosphere and ionosphereover as Thunder Hi, had two setsof booms of length 5.5 m mounta thunderstorm.One payload had an apogeeof 89 km, and ed orthogonally to each other (and to the spin axis of the the other 154 km. To our knowledge these rockets were the rocket) with a vertical separationof 1.5 m betweenboom sets. first dedicatedto the study of the upward coupling of light- Thunder Hi also carried a fixed bias Langmuir probe operated ning and thunderstormfrom electricfieldsinto the ionosphere. in the electron saturation region to measure relative electron In addition, a zero-pressureballoon was flown at an altitude densities.The mesosphericrocket (Thunder Lo), which is also of near 25 km, and a rocket-borneparachutepayload with an describedby Kelley et al. [1983], had two boom sets of apogee of •75 km carried electric field sensors aloft to lengths4.0 m and 5.5 m, oriented perpendicularto the spin characterizethe electricfield signaturesat various altitudes in axis along with a verticalprobe of length 1.3 m along the axis. the atmosphere. Both payloads were designedto make dc, low-frequency, and The stratospheric balloon was instrumented with both broadband vector electric field measurements.The highestquasi-stati c vector electric field detectorsand a broadband frequencyresponsefor one componentof the electricfield was VLF electricfield detectorwith a maximumfrequency of 100 48 kHz. The responsetime for a full vector measurementwas INTRODUCTION 14 kHz • Schoolof ElectricalEngineering, CornellUniversity,Ithaca,New York. 2 Space SciencesDivision, GeophysicsProgram, University of Washington,Seattle. 3 Departmentof ElectricalEngineering,PennsylvaniaState University,University Park. '• NASA GoddardSpaceFlightCenter,Greenbelt,Maryland. Copyright 1985 by the AmericanGeophysicalUnion. Paper number 4A8386. 0!48-0227/85/004A-8386505.00 9815 on Thunder Hi and 8 kHz on Thunder Lo. The storm consistedof four major cells.Figure 1 showsthe position of eachcell at the launch time of Thunder Lo (August 9, 1981, 0208:00 UT) and 15 s before the launch of Thunder Hi. The locationsof Wallops Island and the stratosphericballoon are indicated along with the trajectoriesof the three rocket payloads.Vertical radar scansof each cell just before launch indicatedheavy precipitationreachingup to at least an altitude of 13 km. Radiosonde data indicated that the tropopause temperature minimum was at 16 km, and we estimate 9816 KELLEYET AL.' ELECTRICAL MEASUREMENTS OVERA THUNDERSTORM, 1 MD thunderstormcampaignis presentedbut with emphasisupon the new and exciting observationsof lightning-relatedtransientsin the ionosphere.A detailed presentationof the "dc" (quasi-static)electricfield data can be found in paper 2. Also found in paper 2 is an analysiscomparing the quasi-static electricfield data with earlier work. A more completetheoretical analysisof the ionosphericresultsis under way and will be presentedin subsequentpublications. EL Balloon DATA PRESENTATION Simultaneous data from the two ballistic rockets, the balloon-borneVLF sensorand a ground-based2-MHz receiver located at Wallops Island, are presentedin Figure 2 for a 6-s interval. The• smoothly varying sinusoidal signals in the top panelsare due to the responseof the rotation dc coupled antenna to the sum of the ambient electric field and the V x B 3• electric field (the electric field induced by the motion of the rocket moving at a velocity V acrossthe magnetic field B). The sinusoidalsignalsare interruptedby nearly simultaneous excursionsof the order of severaltens of millivolts per meter r Hi Thunder measured on both rockets. For this event one rocket was lo- catedat an altitudeof 142km, andtheotherat an altitudeof 88 km. Higher time resolution presentationsshow that the signal at high altitudes is delayed from that detected on the I o I o " I o lower payload by a time consistentwith a propagationspeed 76 76• 75• 75 74. equal to the speedof light. 30' 30 30 The event just prior to 0210:40 consistedof a set of five or Fig. 1. Map showing the location of the four thunderstorm cells and the balloon at the launch time of Thunder Lo. The trajectoriesof six excursionsseparatedby about 0.05 s, each of which was Thunder Lo, Thunder Hi, and the Pennsylvania State University detected on the three airborne platforms and the ground parachute payload(afterdeployment) arealsoindicated. receiver. Such a sequenceis characteristic of the multiple strokeswhich make up a singlelightningflash.The magnitude this to be the location of the cloud tops. It shouldbe noted that there is a positivecorrelationbetweencloud heightand elect, rical activity[Shackford,1960]. In this paper an overviewof the experimentalresultsof the •.o w of the electricfieldpulseat 142km (top panel)due to the first stroke was comparable to the signal detected at 88 km. The signalsat high altitude from subsequentstrokes of the same flash were much smaller than those registeredin the meso- 142 km Thunder Hi oc mmkm Thunder_L.o DC ELECTRICFIELB za km Balloon BROAOBANO o.o -35. o •o.o -40.0 :D .,. O.O I ! I Q km D I I I I I I Grou_nd I I 3 MHzRECEIVER 0.0 I UT 02:10:40 I I 41 I I 42 I TIME I 43 I I I 44 45 Fig. 2. Simultaneousdata from ionosphericrocket, the mesospheric rocket, a VLF receiveron the stratospheric balloon,and a ground-based 3-MHz receiver.Notice the signalsdue to lightningstrokesjust before0210:40UT and at 0210:44.5 UT. KELLEYET AL.: ELECTRICALMEASUREMENTS OVER A THUNDERSTORM,1 Flat -I +0 Plate The Super Arcus parachute-borne payload measured the vertical electric field from ~ 75 to ~25 km, but over a much Balloon Verhcal 28 -046 Balloon + 47 -47 02 Horizontal . ii 00 UT 02 ß 12 O0 ......... .. ß ... Thunder Hi Horizontal 35' Thunder H• Horizontal -351 02 II 20 02'11 30 02 II 50 T•me 02 I:> O0 Time 55 9817 Penn State Vertical 0 02 II 50 02 12 O0 T•me Fig. 3. (Top panels)Direct current electricfield data from the flat plate antenna (located at Wallops Island), the balloon, and Thunder Hi. Three major lightning flashesare indicatedby the arrows and designatedeither intracloud(IC) or cloud-to-ground(CG). The other arrowedeventis due to a relaxationconductivitymeasurementmade by the balloon. (Botton panel) Vertical electric field measurement made by the PennsylvaniaState Universitypayload at the sametime asthe lasteventin the secondpanelfrom the top. longer time span than the other rockets.The dc componentof thesefieldsis treated in paper 2. Sixteentransient eventswere observed,and the first two of these also correspondedto events seen at the higher altitudes. At 0211:19.8 UT a downward electric field transient was seen at 50.5 km, with an amplitudeof 3V/m; and at 0211:54 UT at 47.3 km, an upward field change(cloud-to-groundflash) of 4.5 ¾/m was seen.The rise time of the field changeswas of the order of 0.1 s, but it is uncertainbecauseof telemetryfrequencyresponseand degradation during the flashesthat resultedin an inability to distinguish the individual strokes comprisinga flash. However, after the flash the field persistedfor many seconds,with an initial e-foldingtime of about 1 s and a "tail" of about 7 s time constant.Many of the other transientsalso fitted this pattern, which is describedin more detail by Hale [1983]. It is noted that both of thesetime constantsare much larger than the local atmospheric relaxation time, which from simultaneous conductivity measurementswas determined to be about 0.1 s at these altitudes. The presentationsof Figures2 and 3 do ,aotfaithfully reproduce the temporal behavior of the high-altitude(Thunder Hi and Thunder Lo) transientsbecauseof the long time interval plotted and limited digitization rate (312 Hz) of those telemetry channels. Raw data from higher time resolution channels are presented in Figure 4a. These ac-coupled channelsrepresentthe vector electricfield in the rocket reference frame and had high-passfilters located at 14 Hz with digitization rates of 5 kHz (Nyquist frequencyof 2.5 kHz). Note that thesedata are during the same time span as one of the events in the top panels of Figure 3. Since a band-pass filter is used in theseac-coupledchannels,it is necessaryto take into account the transfer function of the electronics if we wish to analyze the data. Figure 4b showsthe result of applying the inversetransferfunctionto the data. From this presentation we can see the basic characteristicsof the lightningsphere.The individual transient fields lasted for about 10-20 inducedtransientsin the ionosphere.Typically,the signalsrise ms at 142 km and 15-35 ms in the mesosphereat 88 km. To to a peak in lessthan 0.2 ms and have an amplitude of ~ 10 our knowledgethis is the first set of data showingsimulta- mV/m and a period of 10-20 ms. Unfortunately,another set of neousmeasurements of lightning-relatedelectricalphenomena data channelsusing even higher telemetry rates were often extendingfrom ground level into the ionosphere.Another saturatedby the lightningevents,so that it is difficult to deterevent registeredon all four platforms occurred at 0210:44.5 mine an exactrise time. However,it is possibleto determinea and can be seennear the right-handsideof Figure 2. "slew rate" or dE/dr for the initial rise in the transient field Although the balloon broadband receiver detected the using thesechannels.A typical dE/dt for the initial rise is at sphericassociatedwith the two eventsin Figure 2, no clear least0.03 mV/s, which is equivalentto 30 V/s. dc electricfield changewas registeredon the balloon (some The presentationthus far has been in the rocket reference 100 km horizontal distance).The balloon registeredonly 13 frame. Several of the lightning penetration eventshave been lightningflasheswith dc electricfield changesduringthe flight analyzedin detail for the ionosphericcase.In general(and in of Thunder Hi, althoughthe Thunder Hi payload registered the caseillustratedin Figure 4) the primary componentof the about 70 electricfield "spikes."In fact, there was a frustrating electricfield was vertical and upward. Ground-basedobservaminute or so with very little lightningactivity. When heavy tions do not clearlyidentify all of the lightningflashesseenby lightning activity resumed after 0211 UT, the Thunder Lo the rocket as either cloud-to-ground or intercloud flashes; payload was no longer taking data. Between the times of however,we have associatedthe upward vertical electricfields 0211:00 UT and 0212:00 UT, three very clear major events with cloud-to-groundstrikes.The upward electricfield signawere registeredon the balloon dc electricfield instrument and ture suggeststhat there was a significant component of the on the flat plate antenna. The charactersof these eventsare electric field parallel to the magnetic field, which was also illustrated in Figure 3. The balloon dc electric field data as mostly vertical(dip angleof 70ø) over Wallops Island. Figure 5 shows data from the vector broadband receivers which have well as the flat plate antenna are shown,and insetsare usedto show the data (for 10-s periods)from Thunder Hi and the been transformedto a geomagneticcoordinatesystemusing parachute-borne payload(data are only shownfor the second the attitude information from an on-board gyroscope.From cloud-to-groundstroke).We note that there is very little acthis plot we can see that indeed the parallel electric field is tivity seenfrom the intracloudflashon the high-altitudepay- comparableto and sometimes largerthan eitherof the perpenload while the secondcloud-to-groundflash causesa large dicular components. amount of activity at the higheraltitude. The VLF characteristicsof the lightning event illustrated 9818 KELLEY ET AL.' ELECTRICALMEASUREMENTS OVER A THUNDERSTORM,1 12.0 ward of the storm. The nominal L value associated with the storm was L- 2.6, but the whistler path is associatedwith L - 2.8. The whistlerwas very intensewith a broadband electric field exceeding1.0 mV/m and was accompaniedby other VLF emissionsindicatedby the arrows in the sonogram. It is interestingto note that no whistlerswere detectedby instrumentson the mesospheric rocketor on the stratospheric balloon. In fact, as shown in Figure 7, even though copious numbersof intensewhistlerswere detectedduring the 4-min ionosphericrocket flight, none of thesewere detectedby the receiversbelow the ionosphere.It is considerednormal for undueted whistlersto be detectedabove the ionosphereand 0.0 -12.0 12.0 0.0 -12.0 12.0 not below. However, similarities between ducted whistlers de- tected at Palma, Antarctica Oustminutesbefore the launch), '.% and the rocket whistlers shown here indicate . . ß -12.0 02'11'53.8 UT 11.54.0 TIME 12.0 that these were indeed ducted events (D. L. Carpenter, personal communication, 1983). Along with the whistlers, signalsfrom two VLF transmitters, NAA located in Cutler, Maine (17.8 kHz), and NSS in Annapolis,Maryland (21.4 kHz), were also detected[Cornish et al., 1981]. In the gray-scaleformat of Figure 6 both VLF transmittersappear as horizontal lines. In the stratosphere (not shown) and in the mesospherethe trasmitter signalsare linearly polarizedas expected.In the bottom panel of Figure 6 this can be seenas a modulation in intensityas the spacecraft spins;that is, the horizontal line correspondingto the transmitter signal is modulated. However, in the ionospherethe transmittersignalsare circularlypolarized,and no spin modulation is observed.A similar phenomenonwas reported by Kintner et al. [1983] based on data from rockets flown over the Siple transmitterin Antarctica and clearly showsthe transitionsfrom "vacuum"propagationto the whistlermode. • ..'..';%,,½ :/,:./..,.•..v•.,.,•,.:,,,,., 0.0 ß . .., -12.0 12.0 ß -12.0 DISCUSSION 12.0 Direct o. o '*"'"'. •'*••;•!.•".. . -12.0 02:11:53.8 UT 11:54.0 Current Electric Fields and ConductivityMeasurements This sectionbriefly reviewsthe resultsdescribedin paper 2 as they are relevant to the overall experiment.The stratosphericballoon data, which were continuousbeginningnearly TIME SOUTH 12.0 Fig. 4. Electric field data from the ionosphericrocket (Thunder Hi). The three panelscorrespondto three orthogonaldirections:two perpendicularto the rocket spin axis and the third along the spin axis. (a) Raw data from Thunder Hi. Notice the electricfield transientswhich are inducedby lightning strokes.(b) Electricfield data that have been corrected(phaseand amplitude)to accountfor the effectsof the band-passfiltersusedin the data channels.Notice that the lightning-induced electricfield signalsrise rapidlyto a peak and then decayin an exponentialmanner. -12.0 EAST 12.0 0.0 -12.0 previouslyin Figure 2 are shownin the frequency/timespectrogram of Figure 6. The plot is a gray-scalerepresentationof the intensity and frequencycharacteristicsof the signal as a function of time detectedin the mesosphere(bottom panel) and in the ionosphere(top panel). Each of the five strokesis seen as a broadband signal in the two detector outputs, a signaturecharacteristicof lightning-generatedVLF measurementsmade in the atmosphereand termedspherics.On Thunder Hi, an intensesubprotonospberic (SP) whistlerwas detected along with an intenseclassicalwhistler.The delay time and nose frequencyof the classicalwhistler indicated a propagation path to the southernhemisphereand back slightly pole- PARALLEL 12.0 '• 0.0 ß -12.0 02-11' 53.8 UT 11' 54.0 TIME Fig. 5. Ionospheric electric field data which have been transformed into geomagneticcoordinates.Notice that the lightningtransientshave a significantelectric field parallel to the earth's magnetic field. KELLEYET AL.' ELECTRICAL MEASUREMENTS OVERA THUNDERSTORM, 1 9819 9820 KELLEY ET AL.: ELECTRICALMEASUREMENTS OVER A THUNDERSTORM,1 o Oo KELLEYET AL..'ELECTRICALMEASUREMENTS OVER A THUNDERSTORM, 1 9821 2 hours before the rocket flights, indicated that the rockets were launched into the waning stagesof a seriesof thunderstorm cellswhich startedshortly after midnight on August9, 1981. The dc electricfield in the stratosphereat a distanceof nearly 100 km pointeddirectly away from the most activecell, the storm cell which the rockets passedover (see Figure 1). The conductivity measurementson the balloon and bluntprobe payloadswere lower than thoseusedin earlier theoretical studies.This stronglyaffectsthe amplitude predictionsof electricfield mapping into the ionosphere.Using the conductivity and electricfield measurements, it is shown in paper 2 ever, that the ionosphericfields had a significantcomponent parallel to the magneticfield. The classicalJoule heating rate at 60 km over the storm and 40 km from the storm zenith. energyinputwill be muchlessthan o'oE•, • because of the aoE•, 2 forsucha transient fieldisquitelarge,andthenetheat input very significant,even though the temporal duration of the field is short. However, we agree that it is likely that the heatingrate is actuallymuchlessthanaoE•, :. Assuming a mean free path of 0.2 km for 100 km altitude, an electron in a 10-mV/m electric field will be acceleratedto approximately 5 times the electronthermal velocity beforesufferinga collision. Hence it is likely that a collectiveinstability will develop becauseof the high electron drift velocities.The collectiveinstathat verticalcurrentdensities as highas 120pA/m2 wereseen bility will causean anomalousresistivityand implies that the This is well above the nominal "electrosphere"referred to in reduced conductivity [Siefring and Kelley, 1983]. We point out in passing that laboratory experiments dealing with many atmosphericelectricalstudies['cf.Chalmers,1967]. It should be noted that the measured conductivity (see anomalous resistivity often induce the desired parallel electric Figure 3 of paper 2) is somewhatdifferent from the conduc- field component by dischargingan external coil I-cf.Hambercreatesan electrotivities used in theoretical calculations by Park and Dejna- ger and Jancarik, 1971]. The resultingt?B/t?t karintra [1973] (on the mapping of dc thunderstorm electric magneticpulse not unlike the transientsignal associatedwith fields)and by Dejnakarintraand Park [1974] (on the propaga- lightning. The perpendicular componentsof the electric field trantion of lightning-inducedelectricfields).The theoreticalconductivity and the measuredconductivityare equal at about 33 sientsmay also have important effectson the ionosphere.One km; however,they differ by 2 ordersof magnitudeat 70 kin. possibleimplication has already been treated [Kelley et al., The conductivityscaleheight usedby Park and Dejnakarintra 1984] in the context of the explosive spread F phenomenon was 6 km while the experimentaldata show a scaleheight of 8 first reported by Woodmanand LaHoz [1976] and more rekm below 40 km and 11 km above 40 kin. The conductivity cently studied by Woodman and Kudeki [1984]. The latter and the conductivityscaleheight are very important in deter- paper shows a good correlation between the rapid onset of mining how the dc and transient electric fields map in the 50-MHz backscatterfrom the F region and the detection of spherics.Kelley et al. [1984], using propertiesof the transient atmosphereand ionosphere. Direct current electric fields of the order of 0.1 V/m were fieldsreported here, showedthat a modified two-streaminstameasuredby the balloon at a distanceof nearly 100 km. This bility should operate and result in a 10- to 30-dB increasein the 3-m (50-MHz) backscatter signal strength within some is somewhat surprising since Park and Dejnakarintra [1973] tens of milliseconds(a characteristic of explosive spread F). predictfieldsof lessthan 10-: V/m for an averagethunderstorm at this distance.It may be that the differencesbetween These data and theoretical studiesmay also explain data presentedby lsted [1954] which showeda relationshipbetween these data and the theoretical calculations are due to the conductivity model usedin thesecalculations.It is also interesting lightning and ionosphericdisturbances. Detailed theoreticalcalculationsof upward lightning penethat the Pennsylvania State University payload detected dc storm-relatedfields up to an altitude of 70 km. More detailed tration do not yet exist at the frequenciesof interest here. discussionof the dc electricfields measuredduring the storm Dejnakarintraand Park [1974] studiedthe upward couplingof electric fields in the frequencyrange from dc to 10 Hz. Burke can be found in paper 2. [1975] has also made calculationsin the ULF range. Although the data are not directly applicable to the full freElectric Field TransientsDue to Lightning quency range of the present observations,we can conclude It is difficult to make a comparisonbetween the lightning that the data are in qualitative agreementwith the theory in transientsmeasuredin the ionosphereand mesosphereand the the following sense.The calculationsshow an attenuation of transients measured in the lower atmosphere.In the lower the field strength which is quite severeat dc but which deatmospherethe recoverytime for the electricfield was longer creaseswith increasingfrequency.The dc field measurements than the time betweenlightning return strokes.Thus the low- made in the ionosphereon Thunder Hi indicate that the value frequencyfield changesseenin the lower atmospherewere due is lessthan the instrumentthresholdof a few millivolts per to the entire lightning flash. This is different from the iono- meter [Holzworth et al., this issue].This is consistentwith the sphericand mesospheric caseswherewe have found that dis- theoreticalestimates,sinceDejnakarintraandPark [1974] pretinct field changesare associatedwith each return stroke. We dicted an ionosphericdc field of lessthan 1 mV/m for a large should note that it is possibleto measure the electric field thunderstorm cell while Burke's [1975] estimates were even responseto each stroke in the lower atmosphereand on the lower. ground.However,the frequencyresponseof our instruments did not allow this. So far, we have not attempted to identify the higher-frequencyelectricfield signaturesin the ionosphere and mesospherewith any of the severalwell-studiedlightning processes(e.g., the step leader), as Beasleyet al. [1982] and othershave done for ground-basedobservations.However, we do plan to pursuethis in the future. The data presentedhere show the first observationsof transient electricfield penetrationinto the ionosphere.At 150 km altitude the verticalcomponentof the transientswas primarily upward. This in itselfis not surprising.It is remarkable,how- VLF Measurements In this experimentwe can directly comparethe intensity of the sphericin the ionospherewith the whistler it creates.Furthermore, we can perform this comparison at virtually the same altitude and in virtually the same plasma conditions. Such a comparisonrevealsthat the signalstrengthsare similar over the entire frequencyrange of the whistler (i.e., 1 kHz to 8 kHz), even though the whistler wave packet has traveled some 30,000 km. The peak of the intensity for the returning whistlers was always in the frequencyrange of 4-6 kHz. In this 9822 KELLEYET AL.: ELECTRICAL MEASUREMENTS OVERA THUNDERSTORM, 1 frequency range the whistler was typically greater than its associatedsphericintensity by 10 to 15 dB. To get a conservative estimate of the lossesalong the path (i.e., from the northern hemisphereto the southern hemisphereand back) we cal- servedfor NAA, NSS, and other VLF transmitters[Imhof et al., 1983; Goldberget al., 1983]. culate the collisional attenuation of the ¾LF waves, but we out the launch of three rockets and a balloon over an active thunder- ignore path lossesdue to spreading of the VLF waves. The storm relied upon the dedicationand technicalexpertiseof a number of individuals.These includeD. Detweiler and R. Long of the Wallops Flight Center, who managed the Cornell payloads, M. Silbert, who was the balloon launchmanager,and J. Brown, who handled the Super Arcus launches.We would also like to thank D. L. Carpenter for his helpful insightsin interpretingthe whistler data. The work at Cornell University was sponsoredunder NASA grant NSG-6020, at University of Washington under NASA grant NAG5-604, and at PennsylvaniaState University under NsG-6004. The Editor thanks the refereefor his assistance in evaluatingthis attenuation is calculated in a manner similar to Helliwell [1965].We usean E regiondensityof 2 x 10'• cm-3 (asmeasuredby the rocket) and an F region model [Helliwell, 1965] with foF2 of 7.5 MHz (as measured by an ionosonde at Wallops Island). We assumea reflectionheight of 95 km in the southern hemisphere and no attenuation above altitudes of 1500 km. This analysisyields an attenuation which is proportional to the squareroot of the frequencyand is equal to 7.4 dB at 4 kHz. The experimental observationsthus indicate that at least a 17- to 22-dB amplification must have occurred along the path (in the 4- to 6-kHz range). These whistler observationshave also given rise to a new model for generationof whistlersin situ above a thunderstorm cell. The interpretationwe presenthere and plan to quantify in a future publicationis that the area of the ionosphereabove an active thunderstormcan be viewed as an aperture antenna. The spatial and temporal variations in electric field on the aperture are prescribed by the penetrating transient electric fields from the lightning discharges.This antenna generates waveswhich propagatein the whistlermode with a variety of wave normal angles.If a duct is illuminated at the base of the protonosphere,it will trap some of the energy,guide it to the other hemisphere,and provide a return path for the signal. The returing signal is in the whistler mode and hence,because of its relatively short wavelength,will not be readily transmitted acrossthe baseof the ionosphere.In this model there is no need to make a "leaky waveguide"sourcefor magnetospheric whistlers,sincethey are producedin situ in the ionosphereby the transient fields we report here. The VLF emissions which "surround" the classic whistler Acknowledgments.The successwith which we were able to carry paper. REFERENCES Beasley,W., M. A. Uman, and P. L. Rustan,Jr., Electricfieldspreceding cloud-to-groundlightning flashes,d. Geophys.Res., 87, 4883, 1982. Burke, H. K., Large scale atmosphericelectricfields:•Comparisons with balloon data, Ph.D. thesis,Rice Univ., Houston, Tex., 1975. Chalmers,J. A., Atmospheric Electricity,Pergamon,New York, 1967. Cornish,C. R., R. F. Pfaff, S. Powell, P. Kintner, M. C. Kelley, and R. Holzworth, Simultaneous measurements of VLF transmissions in the mesosphere and stratosphere (abstract), Eos Trans. AGU, 62(29), 592, 1981. Dejnakarintra, M., andC. G. Park,Lightning-induced eleciricfields in the ionosphere,J. Geophys.Res., 79, 1903, 1974. Goldberg,R. A., S. A. Curtis,J. R. Barcus,C. L. Siefring,and M. C. Kelley, Controlled stimulation of magnetosphericelectrons by radio waves:Experimentalmodel for lightningeffects,Science,219, 1324, 1983. Hale, L. C., Experimentallydeterminedfactorsinfluencingelectrical coupling mechanisms,in Weather and Climate Responses to Solar Variations, edited by B. M. McCormac, p. 309, Colorado Associated University Press,Boulder, Colo., 1983. Hale, L. C., C. L. Croskey, and J. D. Mitchell, Measurementsof middle-atmosphereelectricfields and associatedelectricalconductivities, Geophys.Res.Lett., 8, 927, 1981. Hamberger,S. M., and J. Jancarik,Dependenceof "anomalous"conductivity of plasma on the turbulent spectrum,Phys. Rev. Lett., trace (Figure 6) are also of great interest.The higher-frequency 25(15), 999, 1971. emissionsindicated by arrow 1 are near the local lower hybrid Helliwell, R. H., Whistlersand RelatedIonosphericPhenomena, pp. frequency and may have been generatedby a parametric in61-72, Stanford University Press,Stanford, Calif., 1965. stability [Riggin and Kelley, 1982]. The lower-frequencyemis- Holzworth, R. H., High latitude stratosphericelectricalmeasurements in fair and foul weather under various solar conditions, J. Atmos. sions(arrow 2) are similar to signalsobservedto accompany Terr. Phys.,43, 1115, 1981. one-hop fixed frequencyemissionsfrom the Siple transmitter [Kintner et al., 1983] which are explainedby nonlinear inter- Holzworth, R. H., and Y. T. Chiu, Sfericsin the stratosphere,in Handbookof Atmospherics II, editedby H. Volland,pp. 1-20, CRC actions with magnetosphericparticles near the equatorial Press, Boca Raton, Fla., 1982. plane. If true, this would imply that particle precipitation is Holzworth, R. H., M. C. Kelley, C. L. Siefring,L. C. Hale, and J. D. Mitchell, Electricalmeasurements in the atmosphereand the ionoinducedby the thunderstormactivity. This is consistentwith sphereover an active thunderstorm,2, Direct current electric fields the wave amplification discussedabove. However, it is also and conductivity, J. Geophys.Res.,this issue. possiblethat more than one duct was presentand that these Imhof, W. L., J. B. Reagan,H. D. Voss, E. E. Gaines, D. W. Datlowe, emissionsare due to leakagebetweenducts. J. Mobilia, R. A. Helliwell, U.S. Inan, J. Katsufrakis, and R. G. Joiner, Direct observationof radiation belt electronsprecipitated We can compare the amplitude of the transmitter-induced by the controlled injection of VLF signalsfrom a ground-based ¾LF signalsto the amplitude of the broadband VLF signals transmitter,Geophys.Res.Lett., 10, 361, 1983. generatedby the lightning strokes.The NAA transmitter sigIsted, G. A., Irregularitiesin the E region causedby atmospheric nal had an amplitude of •--0.3 mV/m in the ionosphere, and electricity,in Reportof thePhysicalSocietyConference on thePhysthe NSS transmitter signal had an amplitude of •--1.0 mV/m. icsof the Ionosphere,p. 105,PhysicalSociety,London, 1954. The lightning spherics(VLF emissions)varied greatly in am- Kelley, M. C., C. L. Siefring,and R. F. Pfaff, Large amplitudemiddle atmospheric electric fields: Fact or fiction?, Geophys.Res. Lett., plitude, depending on the stroke, and ranged from 1 to 10 •0(8), 733, 1983. mV/m over the entire frequency range below 24 kHz (the Kelley, M. C., D. T. Farley, E. Kudeki, and C. L. Siefring,A model for maximum frequency our instruments could detect). This equatorial explosivespread F, Geophys.Res. Lett., 11(12), 1168, 1984. implies that the VLF power deliveredto the ionosphereby a sphericmay be as much as 100 times greaterthan that of the Kintner, P.M., R. Brittain,M. C. Kelley,D. L. Carpenter,and M. J. Rycroft,In situmeasurements of transionospheric VLF waveinjecNSS transmitter, but of course the power is spread over a large band of frequencies and is only present for a short period of time (a few milliseconds).Thus lightning may cause particle precipitation in the same manner as has been ob- tion, J. Geophys.Res.,88, 7065, 1983. Mitchell, J. D., D.C. Schoder,K. J. Ho, K. Domagalski,and R. O. Olsen,Subsonicprobe measurements of middle-atmosphere electrical parameters,AIAA J., 20, 503, 1982. KELLEYET AL.' ELECTRICALMEASUREMENTS OVER A THUNDERSTORM, 1 Mozer, F. S., and R. Serlin, Magnetosphericelectric field measurementswith balloon, J. Geophys.Res.,74, 4739, 1969. Park, C. G., and M. Dejnakarintra, Penetration of thundercloudelectric fields into the ionosphereand magnetosphere,1, Middle and subaurorallatitudes,J. Geophys. Res.,78, 6623, 1973. Riggin,D., and M. C. Kelley, The possibleproductionof lower hybrid parametricinstabilitiesby VLF ground transmittersand by natural emissions,J. Geophys.Res.,87, 2545, 1982. Shackford,C. R., Radar indicationsof a precipitation-lightningrelationship in New England thunderstorms,J. Meteorol., 17, 15, R. Green, M. C. Kelley, P.M. Kintner, M. Larsen, R. F. Pfaff, and C. L. Siefring,Schoolof ElectricalEngineering,Phillips Hall, Cornell University, Ithaca, NY 14853. L. C. Hale and J. D. Mitchell, Department of Electrical Engineering, PennsylvaniaState University, University Park, PA 16802. R. H. Holzworth, Space SciencesDivision, GeophysicsProgram, University of Washington,Seattle,WA 98195. D. Le Vine, NASA Goddard Space Flight Center, Code 913, Greenbelt, MD 20711. 1960. Siefring,C. L., and M. C. Kelley, The transientelectricalresponseof the ionosphereto lightning strokes(abstract),Eos Trans. AGU, 64, 778, 1983. Woodman, R. F., and E. Kudeki, A causalrelationshipbetweenlightning and explosivespreadF, Geophys. Res.Lett., 11(12),1165,1984. Woodman, R. F., and C. LaHoz, Radar observationsof F region equatorialirregularities,J. Geophys.Res.,81, 5447, 1976. 9823 (ReceivedOctober22, 1984; revisedFebruary 12, 1985; acceptedApril 18, 1985.)