PREPRINT Characteristics of cloud to ground lightning flashes over Sweden Upul Sonnadara1,*, Vernon Cooray2 and Thomas Götschl2 1 Department of Physics, University of Colombo, Colombo 3, Sri Lanka 2 Division for Electricity and Lightning Research, Uppsala University, Sweden Abstract – A detailed study on the characteristics of cloud-to-ground lightning flashes over Sweden was carried out for the period 1987 to 2000 using the data obtained from the Swedish lightning locating network. Results are presented by analysing over half a million lightning ground flashes. The average variation of the annual mean of the total number of flashes was found to be 37%. About 12% of the cloud to ground flashes were positive flashes and their average variation about the annual mean was 26%. The average peak currents were fairly constant over the years with variations as little as 4% for negative flashes and 5% for positive flashes. The average peak current values for negative and positive flashes were –29.90 kA and +63.97 kA respectively. A correlation between the mean monthly flash count and percentage of positive flashes was seen. A similar relationship is seen with the mean monthly flash count and the peak currents for both polarities. In general, high flash density and high peak currents were observed in the southern part of Sweden where most of the major cities are located. Flash densities exceeding 0.4 flashes km-2 was observed for several large cities. The maximum flash rate of 32 flashes hr-1 within a 10 km radius was seen in Jönköping (14.18° E, 57.78° N) in the Province of Småland. PACs: 52.80.Mg 1. INTRODUCTION Characteristics of cloud-to-ground (CG) flashes in Swedish thunderstorms have been studied in the past by many researchers [1-7]. Some of the earliest work initiated by Norinder dates back to the late 1950’s [1]. Most of these studies were carried out remotely either by using lightning flash counters or by using a sensor system such as flat plate antenna to sense the electric field generated by distant lightning strikes and then digitizing the detected signals using transient recorders. Over the years, these studies have provided valuable insight into the understanding of the physics of the lightning process as well as to understanding the many engineering problems caused by lightning generated electromagnetic fields. However, it should be noted that detailed analysis on the individual lightning waveforms could be carried out only by capturing a very small sample of events in a given thunderstorm due to the limitations in the digitising hardware. In addition, most of the studies the exact strike location of lightning is not known and very often the distance is estimated from the time to thunder measurements.
*
Corresponding author. Tel/Fax: +9411-2584777; Email:
[email protected]
PREPRINT Today, lightning locating systems are used in many countries to study long-term characteristics of CG lightning flashes. These systems have a typical range of 600 km, with many sensors interconnected to form lightning detection networks that span thousands of kilometres that can capture lightning in vast land masses [8-12]. However, they are more suitable for studies that involve producing average lightning parameters and distributions since they are weak in extracting the details of individual waveforms but strong in detecting and processing data from many different thunderstorms. It should be noted that to produce average lightning parameters, these systems should be operated over many years and thus special attention should be paid to the frequent changes in network configurations. In recent years, there have been several attempts to study the long term characteristics of Swedish thunderstorms using the data provided by the Swedish lightning locating system [1314]. However, comprehensive studies are not yet available in the literature that could provide an insight into the climatology of lightning characteristics over Sweden. The purpose of this work was to analyze the CG lightning flash characteristics over Sweden using the data provided by the Swedish lightning detection network. During the selected time period (1987-2000) the network did not undergo any major changes that could affect the results presented here. The data were restricted to study the flash characteristics over the Swedish landmass. The results are presented for the flash densities, percentage of positive flashes, lightning peak current amplitudes and flash rates. 2. DATA In late 1979 a lightning location system was installed in Sweden by the Institute of High Voltage Research, today Division for Electricity and Lightning Research of Uppsala University. The system was a so-called magnetic direction finding system manufactured by the Lightning Location and Protection (LLP), which later became part of Global Atmospherics (GA). Today GA is owned by VAISALA, a Finnish company. At the beginning the system had two sensors, but soon it was expanded to four sensors, model DF 80-02, and one processing unit PA. Later the system was further expanded to include 8 direction finders with an upgraded model ALDF-141. The old processing unit was also replaced by two new Advanced Positioning Analysers, APA. In 1987 operation of the system was taken over by state power system operator Svenska Kraftnät but data on a real time basis was being stored at the Institute of High Voltage Research. In 1992 the system was expanded to 9 sensors by placing a station in Kiruna. In early 1997, the sensor in Luleå was replaced with the new IMPACT model and one processing unit was upgraded to APA 283 T. The system could not take advantage of the new IMPACT technology because it requires two or more sensors to operate correctly, so the new IMPACT sensor operated as an old DF sensor.
PREPRINT In spring of 2000, one sensor was moved from Uppsala to Västerås. In autumn of 2000, the old APA was replaced by a new computer based positioning analyser, LP2000. Eight IMPACT-ES stations from the Norwegian system were also added. A bridged connection was also made between Norway and Sweden so that if one LP2000 breaks down, the other could take over. Presentation of lightning records is switched from flash data to stroke data. In early 2002, a joint co-operation agreement was made by Svenska Kraftnät and Swedish Meteorological and Hydrological Institute, SMHI. They decided to split the operation of the system. SMHI upgraded 3 stations to the IMPACT-ES model. Also two new places were chosen, Malung and Umeå. Vitemölla, Visby and Såtenäs were upgraded. One LP2000 positioning analyser was also placed at SMHI. In early summer of 2002 the rest of the Norwegian stations (6) and all stations from Finland (5) were added (IMPACT-ES). Also, 4 stations from the German system were included, one IMPACT-ES and 3 LPATS. One of the old Swedish stations Hudiksvall was taken out of operation. The total number of stations in the lightning location network at present is 33, of which 10 are in Sweden. Figure 1 shows the present configuration of the Swedish lightning locating system. The present analysis is limited to the data from January 1987 to December 2000. During this period, the only major change to the Swedish network was the change from a network of 8 sensors to 9 sensors which was carried out in the year 1992. In this paper, the results are presented without applying any corrections to the detection efficiency of the Swedish network. A detailed study on the detection efficiency of the Swedish network with 8 sensors as well as 9 sensors has been reported elsewhere [13]. Since the bulk of the data used in this study comes from the network of 9 sensors and the efficiency improvement from the 8 sensor network to the 9 sensor network has an impact only in the northern part of Sweden where the lowest activities are reported, the network configuration change has no significant effect on the results presented in this paper. However, the reader should be aware that although the overall detection efficiency of the Swedish network is about 85%, the detection efficiency varies from point to point within Sweden. In addition, all baseline hits that strike on the line joining two stations were not included in this analysis. This was done in order to exclude all hits where the strike locations are not known accurately. The rejection due to this criteria contributed an additional 10% reduction in the efficiency. Thus, the overall detection efficiency of this work is about 75%. Following similar work carried out for the US network [9-11], the spatial resolution of the analysis was defined with a grid having a cell size of 0.2o ´ 0.2o which corresponds to approximately 20 km resolution. However, due to the convergence of latitude lines when moving away from the equator, the cross sectional areas in the grid cells change from 475 km2 to 325 km2. Thus, the analysis was carried out by using the exact area under each cell wherever applicable.
PREPRINT It should be noted that all positive flashes with peak currents less than 10kA were not considered as a valid positive flash by the Swedish network. This is done mainly to avoid misidentification of cloud discharges as weak positive ground flashes [10]. 3. RESULTS A total of 586,392 cloud to ground flashes were recorded during the period 1987 to 2000 within the area bounded by Sweden with a buffer zone of 10km defined with respect to the boundaries of Sweden. The results are organized and presented in terms of flash density, percentage of positives, lightning peak currents and flash rates. Table 1 shows the yearly breakdown of the total data sample. The annual variations of the total number of flashes do not show any visible trend. This observation justifies the argument given earlier regarding no impact on the results presented in this paper due to the changes in the network configuration during these years. For negative flashes, when compared with the annual negative flash average of 37,307, a large increase of 181% and 95% is seen for the years 1988 and 1997 respectively. The remaining years show an average variation of 30% from the annual average. On the other hand, for positive flashes only year 1988 shows a large increase (96%) compared to the annual positive flash average of 4,578. The second largest deviation for positives is seen for the year 1989. However, for positives only a 16% deviation is seen for year 1997 where the second largest deviation is observed for negatives. The reason for the large increase of negative and positive flashes especially during the year 1988 is not understood. The average variation of the total number of flashes during the 14 year period is 43%. The annual variation of the percentage of positives varies from 7% to 18%. The average variation for positive flashes is 26%. The data shows no visible trend in the percentage of positive flashes over the years. The annual mean of lightning peak currents show they are more or less constant over the years. The average peak currents of –29.90 kA and +63.97 kA with a deviation of as little as 4% and 5% is observed for negative and positive flashes respectively. These values agree with the peak current values reported in the literature for early findings but are larger compared to the recent findings which are reported through lightning detection networks in other countries especially for positive flashes [11]. However, these studies show that even other networks had seen similar values initially with a systematic downward trend over the years, which they attribute to the increase in the sensitivity of the detection system in accepting strikes with low signal strengths [11]. Especially, literature shows that with the introduction of the new IMPACT sensors, where the system reports individual strokes rather than the flash, the amplitude values of the average peak currents have reduced by nearly a factor of 2 over the years. The published peak lightning current distributions for first return
PREPRINT strokes sometimes show strokes with peak currents as low as 2 kA. Although it is an interesting question to investigate whether a lightning first return stroke can deliver a current as little as 2 kA, it is beyond the scope of the present analysis to address this issue. As mentioned earlier, the peak current values less than 10 kA for positive flashes were not considered in the present analysis. This was done mainly to exclude misidentified cloud flashes from our analysis. 3.1 Flash density The mean annual flash density values for cloud to ground lighting flashes over Sweden is shown in Figure 2. A buffer zone of 10 km is used beyond the country boundaries to compensate for the location errors in the lightning detection network. The density values were calculated by counting the number of strikes that fall within a set of grid cells bounded by 0.2° ´ 0.2° in longitude and latitude which corresponds to about 20 km spatial resolution, and dividing by the area of the cell. The change in the cell area due to convergence in the latitude lines was accounted for and exact values were used when calculating the densities. In general, high CG flash densities were observed for the southern part of Sweden (see figure 2a). The highest density of over 0.4 flashes km-2 is observed south of Sweden close to Jönköping (14.18° E, 57.78° N). The area bounded by the coastal area in the Southeast of Sweden running from Göteborg upto Malmö and the costal area in Southwest upto Gävle, shows relatively high flash densities of above 0.3 flashes km-2. Highest lightning activity is observed close to the centre of this area. Towards north of this area upto Umeå shows restively low activity. Very little lightning activity is observed in the Northern part of Sweden close to Kiruna, Luleå and Tärnaby. Especially, above latitude 65° the flash densities were below 0.05 flashes km-2. Lightning flash density variation with the latitude has been reported for Sweden in a previous work [14]. The same variation can be clearly seen in this work too. The average flash density over Sweden varies from 0.3 to about 0.03 flashes per km-2 when moving from latitude 54 degrees to 69 degrees. In addition, the mean annual flash density has a strong dependence on the season with summer thunderstorms delivering roughly a factor of 10 or more flashes than winter thunderstorms. In figure 2b the average density distribution for positive flashes is shown. In general, pattern similar to the distribution of all CG flashes was observed for the distribution of positive flashes. For positive flashes, the highest flash density observed is 0.05 flashes km-2 in the southern part of Sweden. In the northern part, above latitude 65° the flash densities are lower than 0.01 flashes km-2. It should be noted that the total efficiency correction for the southern part of Sweden (including the baseline rejections) is roughly 80% where as for the northern part this is
PREPRINT roughly 70%. The numbers reported here should be weighted with the above efficiency factors if one would like to obtain the true ground flash densities. The cell area used in calculating the average densities varies from 325 km2 to 475 km2. However, one may observe slightly higher values for flash densities if the selected cell area is made smaller. 3.2 Percentage of positive flashes The percentage of CG flashes that lower a positive charge to ground is 12% for 1987 to 2000. However, the percentage of positives shows diurnal, monthly as well as geographic variations. Although a lesser number of positive flashes is seen for the northern part of Sweden compared to the southern part, a higher percentage of positive flashes is seen for the northern part compared to the southern part. In the southern part of Sweden the percentage of positives was around 10%. In the northern part the percentage of positives varied from 20% to more than 100% (northernmost region of Sweden). The mean monthly percentage of positive flashes is shown in figure 3a. A large variation is seen over the months with the lowest in July (9.1%) to the highest in January (55.0%). Clearly, the percentage of positives has strong seasonality dependence with a high percentage in winter thunderstorms compared to summer thunderstorms in Sweden. Similar behaviour has been reported elsewhere for mid latitudes [10]. In figure 3b the mean monthly total of all flashes is shown. The highest monthly mean is seen for July with over 17,467 flashes. The minimum monthly mean is seen for the month December with only 10 flashes. Here also seasonality dependence is seen with a higher number of flashes in the summer months which tapers out in the winter months. This is quite the opposite of what was observed for the percentage of positives. We also investigated the relationship between the number of flashes and the percentage of positives by computing means for hourly values (diurnal variations). It was seen that the maximum hourly mean for flash counts occur at GMT time 1500. Although the percentage of positives showed a minimum at this hour (9.6%) the total variation was less than ±3% from the mean. This indicates that the anti-correlation observed with the number of flashes and the percentage of positives more or less has a strong seasonal dependence. In figure 4 we show the correlation between the number of ground flashes and the percentage of positives. The numbers were extracted directly from the monthly mean values shown in Figure 4. A good linear relationship is seen with a correlation coefficient of 0.91 with the fitted results yielding,
d = a - b ´ ln(N ) ………………………… (1)
PREPRINT where d is the percentage of positives, N is the mean monthly flash count and constants a and b having values 57.05 and 5.16 respectively. Closer examination of results presented by Orville and Huffines [10] for monthly flash counts and percentage of positives for cloud to ground lightning observed in Unites States, revealed a similar relationship. A linear fit to their data produced a = 54.61 and b = 3.16. Within statistical errors associated with this process we conclude that on an average both data sets produce the same relationship. 3.3 Peak currents The geographical distribution of the peak current values are shown in figure 5 for negative and positive flashes separately. During the period 1987 to 2000 the Swedish lightning network reported peak currents of only the first return strokes. Thus, the distributions are shown only for the first return stroke peak currents. Except for the top 5% of cases, the average negative peak currents varied from –15 kA to –40 kA over the area covered by Sweden inclusive of a 10 km buffer zone. The higher peak current values are reported in the southern region. Higher negative currents were seen over the costal belt (see figure 5a) as well as over the ocean compared to the land. This was first reported for Sweden in an earlier work by Strandberg [14] which confirms published results by Orville and Huffines [10]. Strandberg has pointed out that since the land and sea surface conditions are quite different, they can lead to different electrical properties of the atmospheric layers above the surface which may lead to formation of space charges that are different over the ocean compared to over land. However, it should be noted that part of this enhancement may have been caused by the limited coverage of the Swedish lightning locating system. The area of the Ocean, including the Baltic Sea as well as the Gulf of Bothnia situated at the fringes of the lightning locating network and only flashes with high amplitudes may pass the trigger requirement in the detection system. This may lead to systematically selecting only flashes with high signal strength. For most of the positive flashes (95%), the average peak current values varied from +20 kA to about +90 kA. Thus, the possible contamination due to misidentified cloud discharges will be less than 5%. In general the peak current distribution of positive flashes showed a similar pattern to that of negative flashes (see figure 5b). We have also looked at the relationship between the observed peak currents and the monthly flash count. A correlation is seen for both negative and positive flashes (see figure 6). In
PREPRINT general high monthly flash counts produced low peak currents. This means, although in the summer months there is high lightning activity over Sweden, the peak current values are lower compared to winter lightning. This behaviour supports earlier published results by Orville and Huffines [10] for United States. They also observed high peak currents for negative flashes in winter months compared to summer months which strongly suggest seasonality dependence. They attribute this to the strength of the electric fields that initiate lightning in winter compared to summer. For tropical weather [12], data collected through a lightning network consisting of LAPTS sensors an opposite relationship has been seen between the same two variables. High peak currents have been observed for negative flashes with high flash counts. No explanation is given for this observation. 3.4 Flash rates The hourly flash rate was computed by counting the number of flashes that strike within a circular area of (10 km radius) within each hour. However, this method will not give the true flash rate since only a portion of the thunderstorm may exist during the whole hour or the storm may last only a fraction of an hour. Nevertheless, this study focuses on only finding out the maximum hourly flash rate in counties of Sweden where there are more than 20,000 inhabitants. Table 2 shows the names of the selected locations, number of inhabitants in the county (län), maximum hourly flash rate together with the maximum flash density. The maximum flash rate was observed for Jönköping with 32 flashes hr-1. Over 20 flashes hr1 was observed for Halmstad, Växjö, Örebro and Karlstad. As expected from the results presented earlier, low flash rates were observed for locations in the north of Sweden. In fact, most of the populated provinces in Sweden are located in the South where the flash rates are high. The diurnal variations show that the maximum flash rates occur during the summer months between 1400 and 1600 hours. In table 2 we also show the maximum flash densities calculated within the same area. A weak correlation is seen between the maximum flash rates and maximum flash densities. The flash density calculated in this section is higher than the values presented in section 3.1 due to the smaller size of the area used in the calculation and the selectivity criteria. 4. SUMMARY AND CONCLUSIONS A total of over 500,000 cloud to ground lightning flashes recorded during the period 1987 to 2000 within the area bounded by Sweden have been analyzed and the results were presented in terms of flash density, percentage of positives, lightning peak currents and flash rates. The main findings of this study are as follows.
PREPRINT 1. The annual variations of the total number of flashes do not show any visible trend. The average variation of the total number of flashes during the 14 year period is 43%. 2. The data also shows no visible trend in the percentage of positive flashes over the years. The annual variation of the percentage of positives varies from 7% to 18%. 3. The annual mean of lightning peak currents show they are more or less constant over the years. The average peak currents of –29.90 kA and +63.97 kA are observed for negative and positive flashes respectively. 4. High CG flash densities were observed for the southern part of Sweden with the highest density of over 0.4 flashes km-2 observed south of Sweden close to Jönköping (14.18° E, 57.78° N). 5. The area bounded by the coastal area in the Southeast of Sweden running from Göteborg upto Malmö and the costal area in Southwest upto Gävle, shows relatively high flash densities of above 0.3 flashes km-2. 6. The mean monthly percentage of positive flashes showed a large variation with the lowest in July (9.1%) to the highest in January (55.0%). Thus, the percentage of positives has strong seasonality dependence with a high percentage in winter thunderstorms compared to summer thunderstorms in Sweden. 7. The observed peak currents and the monthly flash count showed a correlation for both negative and positive flashes with high monthly flash counts producing low peak currents. Thus, in the summer months, although there is high lightning activity over Sweden, the peak current values are lower compared to winter lightning. 8. The flash rate studies showed that low flash rates for locations in the north of Sweden. Most of the populated cities in Sweden are located in the South where the flash rates are high. This study presents the most comprehensive work that has been carried out up to date with the lightning data recorded in the Swedish lightning network. Nevertheless, the results reported for peak current values should not be taken as absolute values since the increased sensitivity in new lightning locating systems has indicated a significant reduction in the peak current values. There are many unanswered questions which undoubtedly require further studies. Especially future analysis that utilizes the data recorded during more recent years (2001-2005) may produce further insight to the lightning climatology of Sweden. Acknowledgements: Financial assistance given by the IPPS of the International Science Programs, Uppsala University (SRI 01/1), and the Swedish Natural Science Foundation for the research grant G-AA/GU 01448-315 are greatly acknowledged. REFERENCES 1. Norinder H., Ark. Geofys., 2(20), 423 (1956) 2. Norinder H. and Knudsen E., Ark. Geofys., 3(18), 367 (1961) 3. Cooray V., and Lundquist, J. Geophys. Res., 87, 11203 (1982) 4. Cooray V. and Lundquist, Radio Sci., 18, 409 (1983)
PREPRINT 5. Cooray V., J. Geophys. Res., 89, 11807 (1984) 6. Cooray and Perez, J. Geophys. Res., 99, 10683 (1994) 7. Gomes C., Cooray V. and Jayaratne C., J. Atmos. Solar Terres. Phys., 60, 975 (1998) 8. Diendorfer G., Schulz W. and Rakov V. A., IEEE Trans. Elec. Comp., 40, 452 (1998) 9. Huffines G. R. and Orville R. E., J. Appl. Metero., 38, 1013 (1998) 10. Orville R. E. and Huffines G. R., Mon. Wea. Rev., 129, 1179 (2001) 11. Orville R. E. et. al., Mon. Wea. Rev., 130, 2098 (2002) 12. Pinto O. Jr. et. al., J. Atmos. Solar. Terres. Phys., 65, 739 (2003) 13. Fernando M, Galvan A., Götschl T., Cooray V. and Scuka V., ICLP, 150 (1998) 14. Strandberg G., UURIE 289-03L, Uppsala University (1993)
PREPRINT Table 1: Summary of the data sample (Swedish lightning detection network): 1987-2000. Year 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
Flashes (Negative) 25,830 105,012 48,548 20,529 22,014 26,707 23,078 39,388 30,486 20,403 72,787 24,303 24,305 38,910
Flashes (Positive) 3,692 8,950 6,905 4,598 3,615 3,804 4,879 3,452 4,213 3,312 5,303 3,056 3,284 5,029
% Positive 0.13 0.08 0.12 0.18 0.14 0.12 0.17 0.08 0.12 0.14 0.07 0.11 0.12 0.11
Peak current (Negative) -31.15 -29.75 -27.27 -28.34 -27.26 -28.34 -31.82 -29.21 -31.10 -29.87 -31.12 -31.00 -31.66 -31.74
Peak current (Positive) +62.44 +58.71 +63.47 +81.03 +60.04 +64.89 +67.24 +65.97 +62.99 +64.32 +58.06 +61.93 +64.95 +59.60
Table 2: Maximum flash rates and flash densities observed for counties in Sweden where there are more than 20,000 inhabitants. The flash rates were calculated within a circular area of 10 km radius of the capital of each county. County Capital Karlskrona Falun Gävle Visby Halmstad Östersund Jönköping Kalmar Växjö Luleå Örebro Linköping Malmö Nyköping Stockholm Uppsala Karlstad Umeå Sundsvall Västerås Göteborg
Population (2002) 149,875 276,636 277,012 57,381 278,551 127,947 327,971 234,627 176,978 253,632 273,412 413,438 1,145,090 259,006 1,850,467 298,655 273,419 255,230 244,319 258,912 1,508,230
Flash Rate -1 (Flashes hr ) 20 16 5 8 28 15 32 7 22 7 25 18 11 7 7 19 21 9 15 20 17
Flash Density -2 (Flashes km ) 0.27 0.15 0.14 0.21 0.23 0.23 0.26 0.09 0.30 0.08 0.15 0.38 0.44 0.13 0.11 0.36 0.46 0.10 0.16 0.23 0.28
PREPRINT
Station Name Västerås Östersund Vilhelmina Luleå Kiruna Vitemölla Visby Såtenäs Malung Umeå
Longitude 16.639 14.505 16.841 22.183 20.423 14.200 18.334 12.707 13.729 20.299
Latitude 59.599 63.196 64.580 65.541 67.852 55.701 57.653 58.437 60.659 63.786
Figure 1: Sensor locations of the Swedish lightning locating network. Sensor at Kiruna was installed in 1992 and the sensors at Malung (replacement) and Umeå was installed in 2002.
PREPRINT
Figure 2: The mean annual ground flash density over Sweden. A buffer zone of 10 km from the country boundary was used to compensate for the errors in the strike location estimates. (a) All flashes (b) Positive flashes
PREPRINT 60
(a) Positive flashes (%)
50 40 30 20 10 0 Jan Feb Mar Apr May Jun Jul
Aug Sep Oct Nov Dec
20000
Number of CG flashes
(b) 15000
10000
5000
0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Figure 3: (a) Mean monthly percentage of positive flashes (b) Mean monthly total of all flashes
PREPRINT
70 y = 57.05 - 5.16 Ln(x ) r = 0.91
Positive flashes (%)
60 50 40 30 20 10 0 1.E+00
1.E+01
1.E+02
1.E+03
1.E+04
1.E+05
Number of CG flashes Figure 4: Correlation between the number of CG flashes and the percentage of positive flashes. Each point represents the average values seen for a given month.
PREPRINT
Figure 5: The peak current distributions (a) negative flashes (b) positive flashes
PREPRINT
100
Peak current (kA)
80
60
40
20
0 1.E+00
1.E+01
1.E+02
1.E+03
1.E+04
1.E+05
Number of CG flashes
Figure 6: Correlation between the number of CG flashes and the peak current. Closed circles: positive flashes. Open circles: negative flashes. Two lines are drawn to guide the eye.
