Ground Detection of Trans-Ionospheric Pulse Pairs by Stations in the National Lightning Detection Network

R.S. Zuelsdorf, C. Casler, R.J. Strangeway, C.T. Russell

Institute of Geophysics and Planetary Physics, University of California, Los Angeles, CA

R. Franz

Los Alamos National Laboratories, Los Alamos, New Mexico

Received October 30, 1997, revised December 18, 1997, accepted December 22, 1997

Abstract. Trans-Ionospheric Pulse Pairs (TIPPs), as detected by the Blackbeard instrument on board the ALEXIS satellite, correlate with signals that the National Lightning Detection Network (NLDN) classifies as "cloud'' lightning with a positive signal polarity (+IC). Correlation is only found for +IC pulses occurring in the 10 ms interval prior to TIPP occurrence. Apart from this single peak, there is no general change in lightning rates around TIPP time. Correlation between TIPPs and +IC strokes is statistically significant with 99.94% confidence. The amplitudes of +IC pulses that are associated with TIPPs are indistinguishable from the amplitudes of pulses that are not. The rise time of +IC pulses correlating with TIPPs, however, does appear to be longer than the noncorrelating +IC pulses, the median value being about 3 times greater than all other +IC pulses. By assuming TIPPs to be generated close to the detecting ground station, we can use the pulse separation time to calculate the source heights for the ground reflection model. The calculated height of TIPPs is consistent with a cloud source.


Since its launch on board the ALEXIS satellite on 25 April 1993, the Los Alamos National Laboratory instrument, Blackbeard, has continually detected paired bursts of radio noise (TIPPs) throughout its lower VHF (28-95 MHz) and mid-VHF (108-166 MHz) ranges. The ALEXIS satellite is in a 70° inclination at an 800 km orbit. TIPPs as detected by Blackbeard have a median pulse duration of 4 µs [Holden et al.,1995] and a median pulse separation of 50 µs [Massey and Holden, 1995]. These signals are dispersed at low frequencies suggesting a source below the ionosphere. As TIPPs are detected at frequencies known to be radiated by cloud electrification processes, it is believed that TIPPs are generated by the electrical activity of clouds [Holden et al.,1995].

In a recent paper [Zuelsdorf et al.,1997] we correlated the geographical location of TIPP occurrences with that of lightning flashes and demonstrated that TIPP production follows the seasonal migration of lightning. However, even though such a general correlation has been established, it has yet to be shown which cloud processes are responsible for TIPP generation. Here we analyze the raw data from individual stations of the National Lightning Detection Network (NLDN) and demonstrate a correlation between TIPPs and a signal which NLDN classifies as "cloud" lightning with a positive polarity. This correlation was first reported by Franz et al., [1996]. We then calculate the occurrence rates of different types of lightning and compare them to the rate of TIPP detection over North America as detected by Blackbeard.

The dual pulses of TIPPs have yet to be explained. The perhaps most promising hypothesis is that the initial pulse follows a direct path to the satellite, whereas the second pulse is reflected off the Earth [Massey and Holden, 1995]. In this paper we attempt to localize the occurrence of a TIPP within Blackbeard's large footprint and show this localization is consistent with the ground reflection hypothesis.

Detection of the Ground Signature of TIPPs by NLDN Ground Stations

We use individual NLDN stations to correlate TIPPs with ground-detected signals. NLDN is owned and operated by Global Atmospherics, Inc. (GAI) and is a network consisting of over 105 sensors. The system is divided between timing only and timing/angle sensors. For this study we utilized data from 68 of the timing-only Lightning Positioning and Tracking System (LPATS) stations. Two of these stations are at the same location as two others, thus giving us a total of 66 unique sites for lightning detection. These stations are spread uniformly across the contiguous 48 states with a median distance to the nearest neighboring station of about 280 km. GAI raw data are categorized based on polarity (positive or negative) and whether or not the discharge is cloud-to-ground (CG) or a cloud discharge. The GAI raw data database consists of time-tagged entries corresponding to the detection of a signal by any one of the stations. One discharge may register at several stations, thus producing several entries. A positive polarity signifies the raising of negative charge or the lowering of positive charge. The peak-to-zero time, the time for the signal to decay from its peak to the zero level, is used to discriminate between CG and "cloud'' discharges. Discharges with peak to zero times less than 10 µs are classified as "cloud'' discharges. The remaining discharges are triangulated to specific cloud-to-ground source locations. NLDN stations utilize a floating threshold. The higher the rate of detection at a given time, the higher the station sets the threshold. For additional information concerning NLDN see Cummins et al., [1995].

The 17 TIPP events had a sub-satellite point between 35 and 50 degrees north latitude and 245 and 275 east longitude. Such a limitation was imposed to ensure TIPP events occurred over the continental U.S., and thus could be detected by NLDN stations. Both the triangulated and raw data were searched to find correlations between TIPP detection and signals detected by NLDN stations. The only significant correlation discovered was that for 7 out of the 17 TIPP events a "cloud'' discharge with a positive polarity (+IC) was detected within the 10 ms prior to TIPP time. One TIPP had two such correlations giving a total of 8 correlating +IC detections. Figure 1a shows the superposed results for .2 seconds on both sides of TIPP time. This figure is dominated by the peak within 10 ms of TIPP time. Figure 1b is the same plot extending out to 10 seconds on both sides of TIPP time and using .5 sec bins. The correlating +IC pulses exhibit no different signal strength or peak-to-zero time than any other +IC entry. The eight correlating +IC pulses, however, appear to have a longer rise time than the noncorrelating +IC pulses. The median rise time for correlated pulses is 2.0 µs, whereas the median rise time for noncorrelated pulses is 0.6 µs. Table 1 provides a comparison between the attributes of correlating and noncorrelating +IC pulses. A total of 68280 +IC pulses were analyzed.


Figure 1. a) Histogram showing +IC correlation with TIPPs within 10 ms of TIPP time. 
Figure 1. b) Histogram showing constant rate in IC lightning within ±10 seconds of TIPP time. 

Table 1.  Comparison of the Attributes of +IC Signals Correlating with TIPPs to Noncorrelated Signals 
Correlated  Range
Rise time (µs)
1.2 -11.6
Peak to zero time (µs)
Signal strength 
(arbitrary units)

Despite the high rate of +IC entries, it is statistically improbable that 7 or more +IC entries would fall in any given 10 ms bin by chance. Seven is used even though we have eight correlating pulses since we believe that this may very well be the detection of one pulse at two separate stations, as discussed below. All 17 raw data files (68280 events) were analyzed in 10 ms bins and the number of +IC detections in each bin were counted. The 17 series were lined up at TIPP time and superposed. Figure 2 is a log plot showing the probability that any given number of +IC pulses would occur in a single 10 ms bin. We have assumed that all +IC pulses are independent of one another, which accounts for the non-Poissonian tail of the distribution. The probability of 7 or more stations detecting a +IC pulse in any 10 ms interval is .06%. Thus we have 99.94% confidence that this correlation is not mere coincidence. This is a most conservative estimate since it regards one of the eight correlating detections as dependent, while regarding all other +IC detections in the database as independent. Alternatively we can use a criterion used below (Lightning Rates) to assume that all +IC pulses detected within 2 ms of an initial pulse are simply the detection of the same pulse at multiple stations and thus dependent events. By eliminating the dependent events, we calculate the probability of obtaining 7 independent +IC pulses in a given 10 ms bin. This gives us a 99.9991% confidence or 1 out of 108000 intervals. Thus we have a range of confidence from 99.94% to 99.9991% that the correlation between TIPPs and +IC pulses is not random.

Figure 2. Logarithmic plot showing the number of times in our 17 superposed 18 minute datasets that a specific number of +IC pulses occur in any 10 ms bin (Total of 108000 bins). 

Beside the 8 coincident +IC pulses, other discharges were detected within the 10 ms prior to TIPP time. Two -IC as well as one -CG were so detected. However the number of these occurrences are so low, that we do not consider them to be statistically significant. The chance of randomly detecting one -CG flash in any given 10 ms superposed bin is about 20%. The chances of detecting 2 -IC pulses randomly is lower but sufficiently probable to be considered an accidental association, approximately 8%. Similarly one +CG flash was detected within the 10 ms prior to TIPP time. The chance of this being a random occurrence is a low 2%. This could be an instance in which a +IC pulse was misclassified as a cloud-to-ground flash. It is believed that some intracloud pulses are longer than the 10 µs criterion and are thus labeled as ground flashes, particularly for positives (M. Uman, personal communication, 1996).

Lightning Rates

Given the correlation between TIPPs and the +IC pulses as detected by NLDN, it is of interest to compare the lightning rates as determined by NLDN with the North America TIPP rate of 0.007 TIPPs per second as detected by Blackbeard (R. Zuelsdorf et al., "Trans-Ionospheric Pulse Pairs (TIPPs): Their Occurrence Rates and Diurnal Variations", submitted to Geophysical Resarch Letters, 1998). We analyze the GAI data corresponding to the 17 TIPP events to calculate both the triangulated and untriangulated CG rates as well as the IC rates within ±10 minutes of TIPP time.

The calculation of rates for IC lightning presents a challenge as there are no triangulated data for IC lightning. Therefore to calculate the rates one must determine whether subsequent detections by stations as listed in the raw database refer to the same IC pulse or are the detection of different discharges. To achieve this end, the raw data databases for all 17 events were analyzed to determine the median time between +IC entries. It should be noted that once a station detects a pulse it is designed to detect nothing for the following 5 ms. Figure 3 is a histogram showing the time between successive +IC entries for all 17 databases. The top curve includes detections from every station. The bottom curve only includes detections which are subsequent at the same station. Figure 3 presents only a small fraction of the distribution. By comparing these curves, it becomes clear that the +IC pulses which occur within 2 ms of each other are recorded by separate stations. (The minimum time between +IC detection at the same station is 2.9 ms). From this distribution it appears that a fraction of the +IC pulses are detected by a separate station within 2 ms of the original detection.

Figure 3. Histogram of time elapsed between subsequent +IC pulses. Note that no two +IC pulses are detected at the same station within 2 ms of each other. 

Under the hypothesis that stations detecting +IC pulses within 2 ms of each other are detecting the same pulse, Figure 4 is a histogram showing the distances between NLDN stations detecting a +IC pulse within 2 ms of each other. The vertical line represents the median distance between nearest neighbors of 280 km. Note that approximately 23% of all stations detecting +IC pulses within 2 ms of each other fall within the 200 to 300 km bin. The median of the distribution is approximately 550 km. This is consistent with the expected disappearance of intracloud lightning beyond the horizon. The direct line of sight to a cloud 10 km high (assuming a 10 km source height) dips below the horizon when the cloud is greater than 360 km away. Stations 550 km apart thus would be able to view directly the same cloud in a lens-shaped region about 170 km centered between them. Thus if we find an association between a TIPP and an IC pulse, we can assume with some confidence that the discharge occurred within approximately 360 km of the station. A further implication of this median distance is that few IC pulses detected by NLDN stations have reached the receiving station via reflection from the ionosphere.

Figure 4. Histogram of the distance between stations detecting +IC pulses within 2 ms of each other. Vertical line represents the median distance between the nearest station in the network. Median distance between stations in the distribution is approximately 550 km. 

An approximation to the actual IC pulse rate is obtained by utilizing the above assumption that all stations detecting a "cloud'' pulse within 2 ms of each other are detecting the same IC pulse. In this method the nine minutes before and after TIPP time for all 17 events are divided into 10 ms bins. All entries in any given bin are then assumed to correspond to a single stroke. Using this method a +IC rate of 3.12 ± .71 pulses/sec was calculated. This is nearly 3 orders of magnitude greater than the rate of detection of TIPPs over North America by Blackbeard. It was also determined that 6.0 ± .8% of all +IC pulses are detected at multiple stations. This is roughly consistent with 1 out of 7 of the +IC detections correlating with a TIPP being detected at two stations. Similarly the -IC rate is calculated as 2.39 ± .42 pulses/sec with 4.6 ± .6% of all -IC pulses being detected at multiple stations.

CG rates are easily obtained from the triangulated data by simply summing the number of flashes and dividing by the time interval. Using the triangulated data, composite CG rates were calculated from all 17 events. CG rates, as well as IC rates, are compiled in Table 2.

Table 2.  Calculated Lightning Rates 
Triangulated +CG discharges .10 ± .01 flashes/sec.
Triangulated -CG discharges 1.3 ± .2 flashes/sec.
Total CG discharges 1.4 ± .2 flashes/sec.
Positive IC rate 3.12 ± .71 pulses/sec.
Negative IC rate 2.39 ± .42 pulses/sec.
Total IC rate 5.51 ± 1.13 pulses/sec.
North America TIPP rate .007 TIPPs/sec.

Testing The Ground Reflection Model

Above it was stated that 360 km provides a good estimate for the range of ground detection of a +IC pulse. Given the spacing of the NLDN stations, this implies that the majority of all IC pulses are only detected at a single station. It is a possibility that the TIPP correlated with detections at two ground stations in actuality is correlated with a single pulse detected at two stations, not with two separate pulses. These two stations were 490 km (1.6 light ms) apart and the signals arrived within .8 ms of each other. Given that the median distance between the nearest neighboring NLDN stations is about 280 km, this is consistent with our observed ground range of IC pulses. Thus it appears that TIPP ground detection is as localized as +IC pulses.

Given the localization of the ground signature of TIPPs, the ground-reflection model can be tested by placing the TIPP event within 550 km (the median of Figure 4) of the detecting station and using the location of the satellite at TIPP time as well as the chirp separation time to determine the height of the TIPP event. This was performed for all eight correlations assuming signals travel at the speed of light. Table 3 lists the altitude of the satellite at TIPP time, its lateral distance from the detecting station, the chirp separation time for the event, and the calculated height of TIPP origin. Notice the height of TIPP origin is consistent with cloud electrification phenomenon. Although this does not prove the ground-reflection hypothesis, it does show our results are consistent with such a model.

Table 3. Calculated Height of TIPP Events 
to TIPP (km)
Height of 
4.8 ± 1.6
12.5 ± 4.4
9.1 ± 3.5
8.0 ± 2.9
8.4 ± 2.9
5.3 ± 1.8
3.7 ± 1.4
8.4 ± 3.1

GAI data suggests that TIPP occurrence is detected by ground stations as a "cloud'' pulse with positive polarity (+IC). Although this correlation has 99.94% statistical confidence, correlation was found for only 7 out of 17 TIPPs. Furthermore the +IC rate over the US of 3.12 ± .71 pulses/sec is much greater than the TIPP rate of .007 TIPPs/sec as determined by Blackbeard. Those +IC entries correlating with TIPP occurrence exhibit no unusual characteristics as compared to all other +IC entries apart from a larger rise time. It remains a mystery as to why the signal strength of the correlated +IC pulse is similar to the noncorrelating pulses, whereas the fields of a TIPP as measured by Blackbeard (on the order of 100 µV/m per kHz of bandwidth, referred to 1 km) are about 10 times those measured by ground observations of cloud processes [Holden et al.,1995]. In a previous paper [Zuelsdorf et al.,1997] we hypothesize that TIPPs may be similar in origin to the positive bipolar pulses detected by Willett et al., [1989] and Shao et al., [1996]. Our ground detection agrees with these signals in polarity. These results are also consistent with the ground reflection model.

Acknowledgments. This work was supported at Los Alamos and UCLA by grant no. STB/UC:95-141 from the Collaborative UC/Los Alamos Research (CULAR) project.

The National Lightning Detection Network is owned and operated by Global Atmospherics, Inc.

We wish to acknowledge useful discussions with Ken Cummins regarding data obtained from NLDN stations.


Cummins, K.L., E.A. Bardo, W.L. Hiscox, R.B. Pyle, A.E. Pifer, NLDN 95: A combined TOA/MDF technology upgrade of the U.S. National Lightning Detection Network, Intl. Aerospace & Ground Conference on lightning and static electricity, Williamsburg, VA, USA, Sept 26-28, 1995.

Franz, R.C., D.N. Holden, R.S. Massey, J.C. Devenport, R.S. Zuelsdorf, R.J. Strangeway, C.T. Russell, A correlation study of TIPPs with lightning activity as recorded by the NLDN, EOS, Transactions, AGU, 77, (46), Fall 1996.

Holden, D.N., C.P. Munson, J.C. Devenport, Satellite observations of Trans-Ionospheric pulse pairs, Geophysical Research Letters, 22 889-892, 1995.

Massey, R.S. and D.N. Holden, Phenomenology of Trans-Ionospheric pulse pairs, Radio Science, 30, 1645-1659, 1995.

Shao, X.M., D.A. Smith, C.T. Rhodes, D.N. Holden, R.S. Massey, J. Lopez, Observations of large-amplitude bipolar electric field charge pulses: Possible sources for TIPP events, EOS, AGU, 77, (46), F87, 1996.

Willett, J.C., J.C. Bailey and E.P. Krider, A class of unusual lightning electric field waveforms with very strong high-frequency radiation, J. Geophys.. Res., 94, 16,255-16,257, 1989.

Zuelsdorf, R.S., R.J. Strangeway, C.T. Russell, C. Casler, H.J. Christian, R.C. Franz, Trans-Ionospheric pulse pairs (TIPPs): Their geographic distributions and seasonal variations, Geophysical Resarch Letters 24, 3165-3168, 1997.

R. S. Zuelsdorf, Institute of Geophysics and Planetary Physics University of California, Los Angeles, 3845 Slichter Hall, Los Angeles, CA 90095-1567. (

C. Casler, 963 Alexandria Dr., Newark, DE 19711

R. J. Strangeway, Institute of Geophysics and Planetary Physics University of California, Los Angeles, 3845 Slichter Hall, Los Angeles, CA 90095-1567. (E-mail:strange@igpp.ucla. edu)

C. T. Russell, Institute of Geophysics and Planetary Physics, University of California, Los Angeles, 3845 Slichter Hall, Los Angeles, CA 90095-1567. (

R. Franz, Los Alamos National Laboratories, MS D466, Los Alamos, NM 87545. (

 (Received October 30, 1997; revised December 18, 1997; accepted December 22, 1997.)

Back to CT Russell's page More On-line Resources
Back to the SSC Home Page