IMPACT−SAFIR COMPARISONS IN FINLAND

Tapio J. Tuomi1

1Finnish Meteorological Institute, Vuorikatu 24, Helsinki, Finland

ABSTRACT: In the summer of 2002, Impact and Safir lightning location systems were in parallel operation inSW Finland. The Impact system was well established while the Safir system still lacked the final adjustmentsin its installation. However, even a preliminary comparison of the two systems seems to confirm the earliersuspicion that part of the weak positive Impact ground strokes are cloud lightning, which is a drawback inimproving the detection efficiency. The comparison also suggests what kind of improvements should beexpected for Safir when it is calibrated into a fully operational system. Especially, the interpretation of locateddischarges as ground strokes should be more cautious.

INTRODUCTIONSince the thunderstorm season of 1998, Finland has been covered by a lightning location system based on

five so−called Impact sensors (Cummins et al., 1998). In 2002, several Norwegian and Swedish sensors werealso connected to improve the location accuracy and detection efficiency of the system. Also in 2002, a newlyinstalled three−sensor Safir system (e.g. Kawasaki et al., 1994), covering a limited area in south−westernFinland, was in test operation.

Safir is called a "total lightning location system", which means that a sensor receives all kinds of VHFpulses and finds their direction of arrival by interferometry. The pulses originate mainly from breakdownprocesses during channel formation, i.e. from cloud lightning and the "cloud end" of ground lightning. Someof the pulses are identified as, or associated with, ground strokes by the output of an LF sensor which issensitive to the return−stroke frequency range.

The installation of Safir was still incomplete, without final sensor alignment and one sensor at a noisy site,and the sensitivity of the LF sensors was not yet calibrated. The system has been further refined after 2002,and the results presented here do not describe the final operational status of the Finnish Safir. The principalaim of this paper is to use the preliminary Safir data as a diagnostic tool for Impact data, but also to find pointswhich may be important in monitoring the quality of data from either type of system. Because of the pre−liminary state of Safir, an overall comparison is not attempted but only certain special cases are considered. Bythis writing, it is also expected that the present manufacturer of both systems, Vaisala Inc., will develop thetwo systems mutually more compatible.

Because of the established status and great number of sensors of the Impact network, its ground strokedetections are considered generally reliable, especially when at least three sensors have detected the stroke.Then, "reliability" means a location accuracy of about one kilometre, the correct polarity, and the properidentification as a ground stroke. However, many Impact detections still involve only two sensors, especiallyin cases where the signal strength is weak. Generally, two−sensor locations are less accurate, and they ofteninclude very weak positive strokes. It is suspected (Cummins et al., 1998) that at least part of the weak positiveground strokes are actually cloud lightning. In these cases Safir may provide help for interpretation. Eventhough the location accuracy of Safir was not adequate at the time, temporal comparison is possible with anaccuracy of 0.1 millisecond, because both systems use GPS timing. The duration of the return stroke is about0.05 ms, so the time accuracy allows in principle resolving a leader from a return stroke, as well as dischargesbetween or after return strokes. Also, the tail of the (electromagnetic pulse of the) return stroke may be longer,especially when detected from a larger distance. Then the main pulse is followed by a zero crossing and anopposite pulse, the overshoot, which may sometimes have a higher amplitude than the main pulse. Then thereis a risk to detect a pulse with wrong polarity and slightly delayed time.

TWO EXAMPLES OF IMPACT AND SAFIR LOCATIONSFig. 1 shows two one−day examples of Impact (left) and Safir (right) ground−stroke locations. The upper

part is June 21st, when a small thunderstorm area approached from SW. The Safir detection area extends toabout 200 km from the SW corner. The lower part (July 4th) had a more violent thunderstorm approachingfrom the south and demonstrates that both systems reach well over Estonia. In this case, Safir found almost asmany strokes as Impact, but in a smaller area. A comparison over the whole summer in the latitude−longitudesquare 59.5−61.5°N, 21.0−26.5°E (the effective Safir area) gives the following Safir/Impact ratios: flashes1.50 and strokes 1.62. That is, Safir detected 50 % more flashes than Impact in the same area, and a flashcontained more strokes (Impact multiplicity 1.60 strokes/flash, Safir 1.74, including both negative and positiveflashes). Although the Impact detection efficiency is not known accurately, it is estimated to be significantlyhigher than suggested by these values. Hence, Safir is believed to have been, before calibration, too sensitivein interpreting discharges as ground strokes. This will be further discussed below. By the way, the averagemultiplicity measured by Impact was unusually low in that season, probably associated with the relativelyscattered nature of the storms. The normal average multiplicity is about 1.8 strokes/flash.

For a closer comparison, it issafest to choose a situation that isnaturally confined to the effect−ive Safir detection area. If thenorth−eastbound trail on June21st is limited to a time before itreaches the Safir detection limit,and the area is also limited in thesouth−east, comparable sets ofImpact and Safir locations areobtained. If a Safir cloud flash isdefined as a group of locatedpoints (pulses) during about onesecond without an associatedSafir ground stroke, the numberof cloud flashes was 257 (notshown). 40 of these had a co−incident Impact ground stroke.The total number of locatedground strokes was 1783, with1459 (82 %) by Safir and 1014(57 %) by Impact; the Safir/Impact ground−stroke ratio wasthus 1.44. But only 610 strokes(34 %) were clearly common toboth systems, and 40 furtherstroke pairs had a time differenceof about 1 ms and/or oppositepolarities. 43 % of the ground−stroke cases were such that bothsystems had detected the samesituation (lightning in the sameregion during the same second)but not just common strokes, and23 % of the cases had nothing incommon. The low percentages ofconsistent data, even when com−paring only times and not locat−ions, reflect the general difficultyin comparisons of different sys−tems, not only the preliminarystate of Safir. (A similar dif−ficulty was faced in the 1980’s incomparisons of flash−counterdata with LLP locations. LLPwas the predecessor of the pre−sent Impact system.)

RELATIONS BETWEEN SUCCESSIVE STROKESBefore going to a comparison involving weak Impact strokes, let us have a look at how successive strokes inmultiple flashes are related with respect to time and polarity. The interstroke interval of a flash is typically 50ms, varying from about 1 ms up to 500 ms (the upper limit set in the Impact system). Data from the wholeseason show that for Impact locations, 6 % of the interstroke intervals are shorter than 10 ms, and 4 % below 5ms; for Safir, the percentages are 17 and 13. Because the most common multiple flash is negative, thepolarities of successive strokes are normally negative−negative (− −). The three other combinations, involvingat least one positive polarity (− +, + −, + +), also occur; their total percentage is 13 for Impact and 22 for Safir.Among them, the (+ +) pair has the lowest frequency. The percentages of short intervals (below 10 ms) withinthe three positive/mixed−polarity groups is 20−30 % for both Impact and Safir. They are much higher than the6 % for the normal (− −) Impact strokes, and also significantly higher than the corresponding Safir percentage.In other words, Safir detected a higher fraction of positive/mixed−polarity flash pairs, and a higher fraction ofits normal−polarity (− −) flash pairs have short interstroke intervals. These two types of differences between

Fig. 1. Two examples of Impact (left) and Safir (right) ground−strokelocations. The map area is 58−68°N, 19−31°E.

Impact and Safir suggest that in its preliminary state Safir was too sensitive in interpreting its detections asground strokes.

WEAK POSITIVE STROKES DETECTED BY IMPACTAs noted above, the data of the 2002 season are not representative enough for a full comparison between

the systems. An interesting limited data set consists of those cases where Impact has detected weak ( < 5 kA )positive strokes. A "case" in this sense includes, in addition to the weak positive Impact stroke, also all otherImpact and Safir detections within about one second (Table 1). No attention was paid to the Safir signalstrengths because of the lack of calibration. Only the polarity was taken into account.

Table 1. The cases where Impact detected weak positive strokes below 5 kA. The other detected activityoccurred within one second from such an Impact stroke.

System Number Impact type Safir type / interpret.

Impact only 168 Ground stroke

Impact and Safir 168 Weak positive Cloud flash

63 Weak positive Pre−stroke discharge

25 Weak positive Interstroke discharge

112 Weak positive Post−stroke discharge

118 Weak positive Positive stroke

60 Weak positive Negative stroke

213 Ground stroke > 5 kA Same as Impact

Safir only 267 Negative stroke

52 Positive stroke

95 Stroke within 1 ms

A total of 546 weak positive strokes were found for which Safir gave the following interpretations: cloudflashes 31 %, positive strokes 22 %, negative strokes 11 %, pre−stroke 11 %, interstroke 5 %, and post−stroke20 %. It is not certain how the three latter categories should be interpreted. In the rightmost column of Table 1e.g. "interstroke discharge" means that Safir reports to have detected two ground strokes and the weak positiveImpact stroke is thought to be some kind of cloud−discharge activity occurring between them, but this is onlya tentative interpretation that cannot be further confirmed with the available data. The time differencebetween Impact and Safir is considered significant, i.e. the weak positive Impact stroke is not assumed to besomething coincident with either of the Safir strokes. It is also possible that the associated Safir strokes are notground strokes (especially because Impact has not detected them) but cloud discharges, as would then alsolikely be the weak Impact stroke. Then the cloud−flash fraction of Safir detections should be higher. Thiscomparison does not represent the Safir cloud/ground flash ratio properly because the selection rejected allSafir cloud flashes which occurred without an Impact recording (within one second). The cloud/ground flashratio was separately studied in the whole Safir data of the season. If a cloud flash is defined as a sequence,within one second, of activity detected by Safir without an LF ground−stroke identification, the number ofcloud flashes turns out to be significantly lower than the number of ground strokes (cf. the June 21st exampleabove), which is against generally accepted statistics, according to which the cloud/ground flash ratio inFinland should be about 2 (Prentice, 1977). Also from this result it is expected that the LF threshold should behigher in an operational Safir.

The interpretation of some weak Impact positive strokes as genuine positive ground strokes is not solelydue to the Safir data. In some cases, one or more weak positive Impact strokes are part of a multiple positiveImpact flash including at least one strong partial stroke. In these cases, it seems quite clear that weak positiveground strokes do exist. So far it is not clear what is the physical lower limit of negative or positive strokes(initiated by downward leaders, i.e. excluding upward leaders from masts and towers), but it may be close to 1or 2 kA.

In addition to the weak positive Impact strokes, the cases also involved other types of strokes. 213 strongerstrokes, of either polarity, were consistently detected by both systems. There were 168 Impact ground strokeswith no detection by Safir, and on the other hand, 267 negative and 52 positive Safir strokes without acoincident or nearly−coincident Impact detection. Furthermore, 95 Safir strokes, not detected by Impact, weretemporally so close (below 1 ms) to other Safir strokes that they were not considered to be separate ground

strokes. Although the Safir ground stroke interpretation was stated above to be uncertain, the opposite, i.e.interpretation into cloud discharges (actual cloud flashes or cloud activity associated with ground flashes) maybe more reliable, that is, true ground strokes are not often disinterpreted as cloud discharges. Hence it mightbe concluded that more than one third (probably more than one half) of the positive Impact strokes below 5kA are cloud flashes and 20−30 % are true positive ground strokes. Some confusion remains between trueground strokes and a closely related cloud−discharge activity, as seen in the existence of the above pre−stroke,interstroke and post−stroke cases (Safir) and in the occurrence of very short interstroke intervals (bothsystems). The fraction of true positive strokes would, of course, increase for amplitudes above 5 kA, butbecause of the uncertainty of the Safir interpretation, let this preliminary comparison with amplitudes below 5kA suffice.

CONCLUSIONWhile it is expected that the manufacturer will sooner or later develop a system that merges data from both

Safir and Impact sensors, it is useful to have a preliminary idea about the different nature of the data theyproduce. An Impact sensor set at high sensitivity will accept not only ground strokes but also some cloudlightning which happens to emit enough power at lower frequencies to be misinterpreted as weak positiveground strokes. On the other hand, an Impact network may also fail to detect genuine ground strokes if theirpulse form is distorted enough with respect to the standards. Additional information from a Safir sensor mighthelp to resolve the contradiction e.g. by providing the VHF component along with the LF signal. It is alsoimportant to acquire a balance between the ground−stroke reports of the Safir LF antenna and the Impactsensor.

The matter of greatest concern at the present stage is the high number of locations detected by one systemand failed by the other, either way, i.e. the low fraction of fully consistent and coincident strokes. This is notevident in the gross similarity of pairs of maps like Fig. 1. After the final calibration of Safir, it is hoped thatthis will be remedied to some degree. It may also be possible to find the distribution of cloud versus groundstrokes as a function of strength in the Impact data and provide the appropriate correction to Impact statistics.

The present comparison cannot give strictly quantitative and generally applicable results of the differencesbetween Impact and Safir systems but rather, guidelines for studies whenever the two types of systems operatein the same area. Even if the manufacturer may provide means to integrate the systems, this may not berealized immediately in all cases, and then studies similar to those presented here should be especiallyvaluable.

REFERENCESCummins, K.L, M.J. Murphy, E.A. Bardo, W.L. Hiscox, R.B. Pyle and A.E. Pifer, A combined TOA/MDF

technology upgrade of the U.S. National Lightning Location Network, J. Geophys. Res., 103, 9035−9055,1998.

Kawasaki, Z−I., K. Yamamoto, K. Matsuura, P. Richard, T. Matsui, Y. Sonoi and N. Shimokura, SAFIRoperation and evaluation of its performance, Geophys. Res. Lett., 21 (12), 1133−1136, 1994.

Prentice, S.A., Frequency of lightning discharges, in: Lightning (ed. by R.H. Golde), Vol. I, Academic Press,1977, p. 487.