Characteristics of unconnected upward leaders initiated from tall structures observed in Guangzhou

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 117,, doi: /2012jd018035, 2012 Characteristics of unconnected upward leaders initiated from tall structures observed in Guangzhou Weitao Lu, 1,2 Luwen Chen, 3 Yang Zhang, 1 Ying Ma, 1 Yan Gao, 1 Qiyuan Yin, 3 Shaodong Chen, 3 Zhihui Huang, 3 and Yijun Zhang 1,2 Received 4 May 2012; revised 5 September 2012; accepted 8 September 2012; published 13 October [1] Forty-five unconnected upward leaders (UULs) occurred in 19 downward negative flashes are analyzed. Each observed UUL is initiated by a downward stepped leader before a new strike point is struck. For each UUL, several parameters are determined when possible mainly by using high-speed images: inception height, inception time prior to return stroke (RS), horizontal distance from the flash s strike point, two-dimensional (2D) distance between the nearest downward leader branch tip and the UUL s inception point at its inception time, 2D length, and 2D average propagation velocity. Their values range from 40 to 503 m (number of samples: 45), <0.1 to 1.32 ms (38), 20 m to 1.3 km (38), 99 to 578 m (21), 0.48 to 399 m (45), and 5.79 to ms 1 (22), respectively. 86% (19/22) of the velocities are smaller than ms 1. No UUL with an inception time prior to RS greater than 0.5 ms is initiated from a structure lower than 300 m. Those UULs with inception heights lower than 300 m seldom exhibit lengths longer than 50 m and only can be initiated by flashes within approximately 600 m, while those higher than 400 m can even reach several hundred meters and be initiated by flashes over 1 km away. The maximum distances for the downward leaders to attract the UULs with inception heights from 100 to 200 m, 200 to 300 m, and over 400 m are approximately 350 m, 450 m, and 600 m, respectively. Citation: Lu, W., L. Chen, Y. Zhang, Y. Ma, Y. Gao, Q. Yin, S. Chen, Z. Huang, and Y. Zhang (2012), Characteristics of unconnected upward leaders initiated from tall structures observed in Guangzhou, J. Geophys. Res., 117,, doi: /2012jd Introduction [2] In response to the approaching downward stepped leader (usually negative) of the natural downward cloud-toground (CG) lightning flash, one or multiple upward leaders can be initiated from protruding grounded objects or from irregularities of the Earth s surface. An upward connecting leader (UCL) is defined as an upward leader that makes contact with a branch of a downward leader finally, and an unconnected upward leader (or discharge, UUL) is defined as an upward leader that fails to make such a contact finally [e.g., Rakov and Uman, 2003]. Although there are a number of still photographs and video recordings that provide the evidence for the presence of UCL or UUL [e.g., Berger, 1 Laboratory of Lightning Physics and Protection Engineering, Chinese Academy of Meteorological Sciences, Beijing, China. 2 State Key Laboratory of Severe Weather, Chinese Academy of Meteorological Sciences, Beijing, China. 3 Lightning Protection Center of Guangdong Province, Guangzhou, China. Corresponding author: W. Lu, Laboratory of Lightning Physics and Protection Engineering, Chinese Academy of Meteorological Sciences, 46 Zhongguancun S. Ave., Haidian District, Beijing China. (wtlu@cams.cma.gov.cn) American Geophysical Union. All Rights Reserved /12/2012JD , 1977; Orville, 1968; Golde, 1973; Krider and Ladd, 1975; Krider and Alejandro, 1983; Lu and Walden-Newman, 2009], there are very few high-speed images that can actually show the propagation characteristics of UCL and UUL under the influence of the downward leader in natural lightning flash. [3] Due to the spatial and temporal randomness of natural downward CG lightning events, it is very difficult to obtain the propagation characteristics of those upward leaders initiated near the ground. Tall structures are usually used for lightning studies because of the high lightning incidence probability. Using their high-speed images of two natural downward flashes struck on two tall structures in Guangzhou, Guangdong, China, Lu et al. [2010] estimate that the twodimensional (2D) length of one UCL is more than 450 m and that of the other is approximately 177 m, and the average 2D propagation velocities of the two UCLs are of the order of 10 5 ms 1. Warner [2010] presents two lightning flashes to a 163 m tower in Rapid City, South Dakota, USA, which exhibit UCLs with lengths of more than 200 m and increasing 2D propagation velocities from 10 4 to 10 5 ms 1. Four downward flashes containing UULs with lengths of several tens of meters are also observed by Warner [2010]. [4] Many experiments have been conducted on the long air gap discharge in the laboratory to study the attachment process of discharge [e.g., Les Renardieres Group, 1977, 1of12

2 1981], which have provided some fundamental data in models. Up to now, although several models have been proposed to simulate the propagation of downward and upward leaders during the attachment process [e.g., Dellera and Garbagnati, 1990a, 1990b; Rizk, 1994a, 1994b; Mazur et al., 2000; Becerra and Cooray, 2008; Arevalo and Cooray, 2009], only one of them preliminarily considers the effect of multiple upward leaders on the attachment process [Arevalo and Cooray, 2009]. Further model research is required to study the interactions between downward leader and upward leaders (including UCL and UUL) and their effects on the attachment process, and more experimental data are required to provide some fundamental information on the natural laws governing the attachment process and to validate the simulation results. [5] This paper focuses on the analysis of UULs from tall structures induced by the downward leader in natural lightning flashes. Several parameters for each UUL are determined when possible from experimental data, including inception height, inception time prior to return stroke (RS), horizontal distance from the flash s strike point, 2D distance between the nearest downward leader branch tip and the UUL s inception point at its inception time, 2D length, and 2D average propagation velocity. 2. Instrumentation and Observation [6] To study the processes of lightning flashes striking on tall structures, a field experiment has been conducted from 2009 in Guangzhou, Guangdong Province, China [e.g., Lu et al., 2010, 2011]. Guangzhou is a metropolis in China which has a lot of tall structures. Our observation site is located on a structure that is approximately 100 m high in the Guangdong Meteorological Bureau. Several instruments are used to simultaneously measure the acoustic, optical, electric and magnetic signals produced by lightning discharges. A photo of the tall structures in the viewing range of our optical observation systems and the plane view of the tall structures higher than 200 m are shown in Figure 1. There are two structures (A and B) that are taller than 400 m, five structures (C to G) between 200 m to 400 m, and many structures between 100 m to 200 m. Some of the structures were constructed during The distance of the tallest structure in the viewing range, the Canton Tower (610 m high in 2009 and finally 600 m high in 2010), from our observation site is approximately 3.3 km, and that of the second tallest structure, the Guangzhou International Finance Center (GIFC, 440 m high), is approximately 2.4 km. [7] A single-station-based three-dimensional (3D) lightning channel imaging system using the differential arrival time of thunder developed by our group [Zhang et al., 2012], which consists of four microphones, is used to record the acoustic signals of nearby lightning. The system can provide information on the distance, orientation angle, and elevation angle of thunder sources. Two Lightning Attachment Process Observation Systems (LAPOS) [Wang et al., 2011] are installed to observe the attachment process occurring near the tops of the Canton Tower and the GIFC. Each LAPOS consists of a lens, a camera body, an optical fiber array, eight photodiodes, and eight amplifiers. A flat-plate fast antenna and a flat-plate slow antenna, with a time constant of 1 ms and 6 s, respectively, are used to obtain the electric field change produced by the lightning discharge. The signals from the LAPOSs, the fast antenna and the slow antenna are recorded by two YOKOGAWA DL750 digital oscilloscopes with a sampling rate of 10 MHz. Several Photron Fastcam high-speed cameras with different performance and configuration were used in our experiment to capture the images of the lightning channels. They are one SA3 with a monochrome sensor, two monochrome SA5s and one color SA5. During our experiment, at least two high-speed cameras operating with two different options are used at a same time: one option is at a sampling rate of ranging from 1,000 to 10,000 frames per second (fps) with a relative large viewing range, and the other option is at a sampling rate of 50,000 fps with a relative small viewing range. The former is used to obtain the overall characteristics of the lightning flashes appearing in the viewing range, while the latter is mainly used to observe the attachment process of lightning flashes striking on the GIFC. One channel of a LAPOS is used as the trigger source for all the instruments. For each triggering event, the GPS time of the triggering signal with an accuracy of 30 ns is provided by a high-precision GPS timing system. [8] In this research, the images observed by the high-speed cameras are mainly analyzed. For the lightning flashes striking on tall structures in the viewing range, their strike points are identified from the high-speed images, their distances are measured using a laser rangefinder, and their heights are calculated from their distances and elevation angles. For several cases with strike points that are outside of the viewing range, acoustic data are used to determine the orientation angles and distances of the main channels of the lightning flashes. The polarity of each RS is indicated from the electric field change wave. The number of strokes striking to each strike point in each lightning flash is identified by combining the high-speed images and the electric field change data. The peak current values of the RSs are obtained from the records of the Guangdong Power Grid Lightning Location System, whose detection efficiency and precision have been evaluated in Chen et al. [2012] by using the observational data of artificially triggered lightning flashes obtained in Conghua, Guangzhou from 2007 to 2011 and natural lightning flashes striking tall structures obtained in Guangzhou from 2009 to Analysis and Results [9] During , a total of 45 UULs were recorded by our high-speed cameras in 19 natural downward lightning flashes. Their characteristics are shown in Table 1. These flashes are numbered as FYYsn, in which YY denotes the last two digits of the year and sn denotes the serial number of a particular flash in all flashes (with or without observed UULs in the viewing range) recorded in the year YY and sorted by time. All of these flashes are negative, six have multiple strokes and only one (F1102) has multiple strike points. These UULs are numbered as UULSN, in which SN denotes the serial number of the UUL in our UUL database. For those multiple UULs in a flash, the UULs are sorted in ascending order of the distances between their inception points and the strike point. Each of the 45 observed UULs is initiated by a downward negative stepped leader before a new strike point is struck, which means that all of them are upward positive leaders. In Table 1, 2of12

3 Figure 1. The tall structures in the viewing range. (a) A photo captured on July 5, (b) The plane view of the positions of the tall structures higher than 200 m. The inset of Figure 1a shows the helicopter pad (440 m high) on the top (432 m high) of the Guangzhou International Finance Center (GIFC, denoted with B). the corresponding RS means the RS produced by the downward negative stepped leader which triggers the UUL. It should be mentioned that sometimes the height of the strike point is not equal to the height of the structure on which the flash struck, because the strike point may be not on the top of the structure, such as tip of its construction crane and somewhere below its top [e.g., Hussein et al., 2007]. Similarly, there also exist difference between the inception heights of some UULs and the heights of the structures where the UULs are initiated. [10] Some high-speed images of eight flashes containing UULs observed during have been presented and preliminarily analyzed by Lu et al. [2011]. Their analysis results are updated and appended in Table 1. In this paper, three flashes are selected to show high-speed images and analysis results in detail Case Analysis F1004: A Flash That Consists of Four Long UULs With Inception Heights Higher Than 200 m [11] A Photron FASTCAM SA5 high-speed camera with a color sensor is used in 2010 and its sampling rate is set at 10,000 fps. F1004, the Case 7 in Lu et al. [2011], is a flash with two RSs occurring in the viewing range of this camera. 3of12

4 Table 1. Characteristics of the 45 Unconnected Upward Leaders Initiated From Tall Structures Observed in Guangzhou During Characteristics of the Corresponding Return Strokes Characteristics of Unconnected Upward Leaders Strike Point Inception Location and Time Description of Observation Data Case Number Beijing Local Time Height Distance Peak Current (ka) Number 2D Length 2D Average Velocity (10 4 ms 1 ) Height Dis1 a Inception Time Prior to RS (ms) Dis2 b Distance 2D Spatial Resolution (m per pixel) Sampling Rate (10 3 fps) F :53: UUL F :00: UUL F :43: UUL F :08: UUL (2D) UUL (2D) UUL (2D) F :13: UUL F :23: UUL < F :18: UUL UUL UUL UUL F :23: UUL UUL UUL F :39: UUL F :44: UUL UUL F :57:15 < UUL F :37: UUL F :52: UUL UUL F :56: UUL (2D) UUL (2D) UUL UUL F :38:45 < UUL27 > UUL UUL F :21: UUL UUL31 > UUL32 > UUL UUL UUL UUL UUL UUL UUL of12

5 Table 1. (continued) Characteristics of the Corresponding Return Strokes Characteristics of Unconnected Upward Leaders Inception Location and Time Description of Observation Data Strike Point Sampling Rate (10 3 fps) 2D Spatial Resolution (m per pixel) Distance Dis2 b Inception Time Prior to RS (ms) Dis1 a Height 2D Average Velocity (10 4 ms 1 ) 2D Length Peak Current (ka) Number Distance Height Case Number Beijing Local Time F :13: UUL UUL F :17: UUL (2D) UUL UUL F :58: UUL a Dis1: horizontal distance from the flash s strike point. b Dis2: 2D distance between the nearest downward leader branch tip and the UUL s inception point at its inception time. Two high-speed images of F1004 captured by the color SA5 are shown in Figures 2a and 2b. The first RS of F1004 occurs nearly at the end of the exposure duration of Figure 2b. Figure 2a shows five upward leaders from three tall structures. One of them is an UCL, which is initiated at a height of 340 m from the tower crane of a 305 m high structure. The others four upward leaders are UULs, numbered as UUL09 to UUL12. UUL09 is initiated at a height of 342 m from the hydraulic crane of a 310 m high structure (the structure D in Figure 1), about 0.4 ms prior to the first RS when the nearest tip of the downward leader branches is about 300 m away in the 2D image. From Figure 2b, it can be determined that UUL10 is also initiated from a tower crane at a height of 239 m, about 0.3 ms prior to the first RS when the nearest tip of the downward leader branches is about 432 m away in the 2D image. The inception point of UUL10 had been obscured by a new 130 m high structure about 520 m away from the observation site since 2011 (Figure 1a). [12] Another SA5 with a monochrome sensor is aimed at the top of the GIFC and is operated at a sampling rate of 50,000 fps. Because the monochrome sensor has higher sensitivity than the color sensor, the high-speed images of the monochrome SA5 show the initiation and propagation of UUL11 and UUL12 better than those of the color SA5. As shown in Figure 2c, both UUL11 and UUL12 are initiated from the helicopter pad on the top of the GIFC. It appears that UUL11 and UUL12 have a same inception point in the 2D image. The inception time of UUL12 is 0.98 ms prior to the first RS, approximately 0.44 ms earlier than that of UUL11. When they are initiated, the 2D distances between their inception points and the nearest tips of the downward leader branches are 578 m for UUL12 and 505 m for UUL F1102: A Flash That Consists of Two Strike Points [13] In 2011, a monochrome SA3 camera was installed and operated at a sampling rate of 1,000 fps; it also has a very large viewing range by using a lens with a focal length of 14 mm. F1102 is captured by this camera. As shown in Figure 3, F1102 has two strike points. Combined Figure 3b and the acoustic data, it can be speculated that the first RS of F1102 likely strikes a 160 m high structure approximately 2.6 km away from the observation site, just out of the left side of the viewing range of the SA3. Only one RS occurs at this strike point. Approximately 144 ms later, the second RS strikes a 270 m high structure approximately 2.4 km away from the observation site (Figure 3c). The downward leader before the second RS propagates along a branch of the downward leader before the first RS with a few branches before it reaches a height of approximately 1.1 km, as shown in Figure 3c, then creates new channels with obvious branches, and finally creates a new strike point. A total of eight RSs strike the second strike point within 558 ms. The distance between the two strike points is approximately 420 m. [14] Both Figures 3a and 3c consist of one UUL. The inception point of UUL21 (Figure 3a) is 356 m high, at a tip of the tower crane of a constructing structure with a final height of 360 m and approximately 600 m away from the first strike point, while UUL22 (Figure 3c) is initiated from the top of a structure, 305 m high and approximately 5of12

6 Figure 2. High-speed images of F1004. Two frames approximately (a) 0.2 ms and (b) 0.1 ms before the first RS, captured by a color high-speed camera with a sampling rate of 10,000 fps. (c) The composite image of 70 frames before the first RS captured by a monochrome high-speed camera with a sampling rate of 50,000 fps. The brightness and contrast of Figure 2 are enhanced for a better view. approximately 18.6 m and two more UULs from the working arm of the crane are observed. A total of four upward leaders, one UCL and three UULs, are initiated from a same tower crane in F1109. During the last 0.1 ms prior to the RS, seven more UULs are initiated from different structures (Figure 4c) less than 0.1 ms prior to the RS. The inception point of UUL30 is located on the top of the structure where the tower crane is installed. UUL37 and UUL38 are initiated from a 110 m high structure 290 m away from the strike point. The final lengths of UUL31 and UUL32 cannot be determined because of the mixture of their paths and the saturated pixels produced by the RS. Only the velocity of UUL33 can be calculated by using the high-speed images, which is approximately ms 1. [16] In F1109, a total of 10 UULs are observed before the RS. This is the maximum number of UULs in a flash in our database. Most of the UULs exhibit weak luminosities, so some UULs channels are denoted by bright points in Figure 4 for clarity Statistical Analysis Inception Height [17] The distribution of the inception heights of the 45 observed UULs in Guangzhou during is shown in Figure 5: seven UULs exhibit inception heights lower than 100 m, and five of them are initiated from the same corner of the structure on which our observation room is constructed (the yellow circle in Figure 1a, 92 m high and 15 m away from the high-speed cameras), e.g., UUL39 in Figure 4c; almost half of the UULs (22 of the 45, approximately 49%) are initiated at heights between 100 m to 200 m; six of the UULs have inception heights between 200 m to 400 m and 10 of them are initiated at heights over 400 m. The large difference between the number and the location distribution of the structures at different height ranges in the viewing range may be the main reason for the obvious difference between the distributions of the number of the UULs in different inception height ranges. No UUL 130 m away from the second strike point. No UUL is observed during the subsequent dart-leader-return-strokes strike the second strike point F1109: A Flash That Consists of 10 Observed UULs [15] Four frames of the high-speed images for F1109 are shown in Figure 4. Obscured by clouds, the channels of F1109 higher than approximately 250 m are invisible. From the brightness distribution of Figure 4c, in which most pixels of the sky are saturated, the RS of F1109 is inferred to occur just at the end of the exposure duration of Figure 4c, and the strong luminosity produced by the RS results in the saturation of all pixels in the high-speed image following Figure 4c. In Figure 4a, approximately 0.3 ms prior to the RS, the UCL and an UUL (UUL33) initiated from the same tower crane are observed. This crane is 520 m away from the observation site, which has a 60 m long working arm and a 15 m counter jib. The inception height of the UCL of F1109 is 145 m. In Figure 4b, the UCL shows a length of Figure 3. High-speed images of F1102 with a sampling rate of 1,000 fps: (a) the last frame before the first RS; (b) the frame containing the first RS; and (c) the frame containing the second RS. In Figure 3a, the brightness and contrast are enhanced and the channel of UUL21 is denoted by bright points for clarity. 6of12

7 Figure 4. High-speed images of F1109 with a sampling rate of 10,000 fps: (a) 0.3 ms; (b) 0.2 ms; (c) 0.1 ms; and (d) 2.0 ms. The channels of some upward leaders in Figures 4a to 4c are denoted by bright points for clarity. Figure 5. The number of the unconnected upward leaders versus the inception height for 45 UULs. 7of12

8 Figure 6. The inception time prior to RS versus the inception height for 38 UULs. initiated from the highest structure in the viewing range, the Canton Tower, is observed in our experiment, which may be due to lack of data, or sometimes obscuration of the tower tip by clouds. All of the 10 UULs with inception heights higher than 400 m are attributed to the GIFC. One UUL, UUL07, is branched from the UCL at a height of 503 m in F0911, in which the UCL is initiated from the tip of the hydraulic crane of the GIFC. UUL01 is also initiated from the hydraulic crane of the GIFC and has the second highest inception height, 487 m. Four UULs are initiated from the helicopter pad on the top of the GIFC and four from the top of the GIFC (Table 1). Guangzhou is a rapidly developing city with many construction cranes (tower or hydraulic) in our viewing range during our experiment. Among the 19 flashes with 20 strike points, four (20%) struck on construction cranes, and 17 (38%) of the 45 UULs are initiated from construction cranes Inception Time Prior to RS [18] In Table 1, the inception time values for seven UULs are not calculated for the following reasons: the inception point is out of the viewing range (UUL02 and UUL03), the luminosity of the leader at its initiation time is indistinguishable (UUL07), or the sampling rate of the high-speed camera is too low (UUL21, UUL22, UUL25, and UUL26, approximately 1 ms exposure duration per frame). The values of inception time prior to RS for 38 UULs vary from less than 0.1 ms to larger than 1.3 ms. There are two groups of the inception time data obtained using the high-speed images with a sampling rate of 10,000 fps (approximately 0.1 ms exposure duration per frame): the values with an accuracy of 0.1 ms, like those of UUL19, UUL40 to UUL45; and the values with an accuracy of 0.05 ms, like those of UUL08 to UUL10, and UUL30 to UUL39. Because the highspeed images cannot provide information on the precise inception time of an UUL and the occurrence time of a RS more accurately than the exposure duration per frame, those values with an accuracy of 0.1 ms in Table 1 have an uncertainty of 0.1 ms. However, some high-speed images, such as those in Figures 2b and 4c, can indicate that the occurrence time of the RS is almost at the end of the exposure duration of a frame or almost at the beginning of the next frame, so some values with an accuracy of 0.05 ms and an uncertainty of 0.05 ms are presented in Table 1, which only consist of the 0.1 ms uncertainty of the inception time of the UUL. For example, the value for UUL31, 0.15 ms, means that its inception time prior to RS is larger than 0.1 ms and less than 0.2 ms, and that for UUL30 is less than 0.1 ms. Those values obtained from the 50,000 fps high-speed images are not processed with similar approach for their lower uncertainty, 0.02 ms. [19] The inception time prior to RS versus the inception height for 38 UULs are shown in Figure 6. It is obviously that the higher the inception height, the earlier the upward leader can be initiated. In our database, no UUL with an inception time prior to RS greater than 0.5 ms is initiated with an inception height lower than 300 m, and no UUL with an inception height higher than 300 m is initiated less than 0.3 ms prior to RS. For those UULs with inception heights lower than 300 m, their values of inception time prior to RS range from less than 0.1 ms to 0.5 ms Two-Dimensional Length [20] The shortest UUL in our database is UUL39 with a length of approximately 0.48 m, and the longest one is UUL26 with a length of approximately 399 m. Figure 7 shows the 2D length versus inception height for 42 UULs. Among the 15 UULs with inception heights higher than 300 m, 12 (80%) are longer than 50 m in the 2D image (six longer than 100 m), and three have lengths between 20 m and 50 m. While among the 27 UULs with inception heights lower than 300 m, only two are longer than 50 m (one longer than 100 m), four are between 20 m and 100 m, and most of them, 22 (81%), are shorter than 20 m. Most (12/14, 86%) of the UULs with 2D lengths longer than 50 m exhibit inception heights higher than 300 m. [21] The absence of UUL with an inception height higher than 300 m and with a 2D length shorter than 20 m or an inception time prior to RS less than 0.3 ms, as shown in Figures 6 and 7, may be due to the weak luminosities of short upward leaders and the long observation distances for those structures higher than 300 m Two-Dimensional Average Propagation Velocity [22] The 2D average propagation velocities of 22 UULs can be calculated from the high-speed images in our database. As shown in Table 1 and Figure 8, the 22 2D average propagation velocities range from 5.79 to ms 1, with six of them being of the order of 10 4 ms 1 and 19 (86%) of them being smaller than ms 1. Wang et al. [2008] observed many aborted upward leaders that occurred from a 100 m high windmill and its 105 m high lightning protection tower in response to nearby lightning, and the velocities of two leaders are estimated to be approximately ms 1, which are consistent with our results Horizontal Distance From the Strike Point [23] Among the 45 UULs, the horizontal distances between the inception points of 38 UULs and their corresponding flashes strike points can be determined using Google Earth. These distances versus the inception heights of the UULs are shown in Figure 9. UUL07, which is branched from the UCL, is not considered here. For the other six UULs without sufficient location information, the 2D distances between their inception points and the strike points are calculated by using the 2D image, which are denoted in Table 1. From Figure 9, it can be seen that the UULs with inception 8of12

9 Figure 7. The 2D length versus the inception height for 42 UULs. heights lower than 300 m only can be initiated by those lightning flashes within approximately 600 m (from 20 m to 510 m for 26 UULs), while the distances from the strike points for the 12 UULs with inception heights higher than 300 m range from 46 m to 1.3 km, and eight (67%) of them are no less than 600 m Two-Dimensional Distance Between the Nearest Downward Leader Branch Tip and the UUL s Inception Point at Its Inception Time [24] The distribution of the 2D distance between the nearest downward leader branch tip and the UUL s inception point at its inception time versus the inception height for 21 UULs is shown in Figure 10. Note that each 2D distance is inferred from a high-speed image by using the spatial resolution of the UUL because the observation distance for the downward leader is unknown. The 2D distance cannot accurately represent the distance in 3D space, especially when the observation distances for the UUL and the downward leader have a large difference. Therefore, the 21 2D distances, ranging from 99 m to 578 m, are approximate underestimations of the actual distances. However, the maximum 2D distance values in the different UUL inception height ranges can provide indications on the characteristics of the maximum distance between the nearest tip of the Figure 8. The distribution of the 2D average propagation velocities for 22 UULs. 9of12

10 Figure 9. The horizontal distance from the flash s strike point versus the inception height for 38 UULs. downward leader branches and the UUL at its inception time. The maximum distance should be approximately 350 m, 450 m, and 600 m for the UULs with inception heights between 100 m to 200 m, between 200 m to 300 m, and higher than 400 m, respectively. 4. Concluding Remarks and Discussions [25] In this paper, the high-speed images consisting of UULs observed in Guangzhou during are analyzed. Several parameters of the 45 UULs occurring in 19 downward negative flashes are determined when possible: inception height, inception time prior to RS, horizontal distance from the flash s strike point, 2D distance between the nearest downward leader branch tip and the UUL s inception point at its inception time, 2D length, and 2D average propagation velocity, whose values range from 40 m to 503 m (number of samples: 45), less than 0.1 ms to 1.32 ms (38), 20 m to 1.3 km (38), 99 m to 578 m (21), 0.48 m to 399 m (45), and 5.79 to ms 1 (22), respectively. Some high-speed images of three flashes are selected to present the occurrence of UULs and to be analyzed in detail. Figure 10. The 2D distance between the nearest downward leader branch tip and the UUL s inception point at its inception time versus the inception height for 21 UULs. 10 of 12

11 [26] Statistical analysis of the characteristics of the UULs shows that: [27] 1. No UUL with an inception time prior to RS greater than 0.5 ms is initiated from a structure lower than 300 m, and no UUL with an inception height higher than 300 m is initiated less than 0.3 ms prior to RS. [28] 2. Those UULs with inception heights lower than 300 m seldom exhibit 2D lengths longer than 50 m and most (81%) of them have 2D lengths shorter than 20 m. Most (86%) of those UULs with 2D lengths longer than 50 m have inception heights of over 300 m. [29] 3. The 2D average propagation velocities of 22 UULs range from 5.79 to ms 1, six of them are of the order of 10 4 ms 1 and 19 (86%) of them are smaller than ms 1. [30] 4. Those UULs with inception heights lower than 300 m only can be initiated by those lightning flashes within approximately 600 m, while those with inception heights higher than 400 m can be initiated by flashes over 1 km away. [31] 5. The maximum distances for the downward leaders to attract the initiation of UULs with inception heights from 100 m to 200, 200 m to 300 m, and over 400 m are approximately 350 m, 450 m, and 600 m, respectively. [32] In our database, several upward leaders can be initiated almost simultaneously from a same structure within a region with scale of several tens of meters, such as those UULs in F1004 and F1007, and UCL and UULs in F1109. Different upward leaders can also be initiated by the downward leader branches of a flash from different structures with distances greater than 1 km. [33] Ten UULs attributed to the GIFC are observed during , while no UUL initiated from the highest structure in the viewing range, the Canton Tower, is observed. Four flashes struck on the Canton Tower are recorded by our high-speed cameras: three exhibit long UCLs from the tip of the Canton Tower; the lightning channel in the highspeed image of the other flash is very blurry affected by clouds and heavy rain. The absence of the UUL with an inception height higher than 300 m and with a 2D length less than 20 m or an inception time prior to RS less than 0.3 ms may be due to the weak luminosities of the short upward leaders and the long observation distances for those structures higher than 300 m. It is thought that the higher the structure, the easier an UUL from it can be initiated. So probably there exist many short UULs from structures higher than 300 m in the viewing range, including the Canton Tower and the GIFC, initiated by nearby lightning discharges, but cannot be identified from our high-speed images. [34] The analysis results of this paper on the characteristics of UULs are hoped to deepen our understanding of upward leaders from tall structures, especially from tall structures complex in urban district. Some characteristics of UULs can help us to study the property of UCL in attachment process, to improve model work, and to validate simulation results. Limited by single station optical observation, several parameters of UULs can be only analyzed in 2D space. As we know, if the lightning channel tilts toward or away from the optical observation system, 2D photogrammetric analyses using the 2D images will have imponderable error. Therefore, multistations optical observations are required to obtain the 3D propagation characteristics of leaders in the future work. [35] Acknowledgments. This work was supported in part by the National Natural Science Foundations of China (grants , , and ) and the Basic Research Fund of Chinese Academy of Meteorological Sciences (grant 2010Z004). The authors would like to thank the three anonymous reviewers for their valuable comments on this paper. References Arevalo, L., and V. Cooray (2009), Influence of multiple upward connecting leaders initiated from the same structure on the lightning attachment process, paper presented at X International Symposium on Lightning Protection, Cent. for Sci. and Technol. of the Non-Aligned and Other Dev. Countries, Curitiba, Brazil. Becerra, M., and V. Cooray (2008), On the velocity of positive connecting leaders associated with negative downward lightning leaders, Geophys. Res. Lett., 35, L02801, doi: /2007gl Berger, K. (1967), Novel observations on lightning discharges: Results of research on Mount San Salvatore, J. Franklin Inst., 283, , doi: / (67) Berger, K. (1977), The Earth flash, in Lightning, vol. I, Physics of Lightning, edited by R. H. Golde, pp , Academic, New York. Chen, L., Y. Zhang, W. Lu, D. Zheng, Y. Zhang, S. Chen, and Z. Huang (2012), Performance evaluation for lightning location system based on observation of artificially triggered lightning and natural lightning flashes, J. Atmos. Oceanic Technol., doi: /jtech-d , in press. Dellera, L., and E. Garbagnati (1990a), Lightning stroke simulation by means of the leader progression model. I. Description of the model and evaluation of exposure of free-standing structures, IEEE Trans. Power Delivery, 5(4), , doi: / Dellera, L., and E. Garbagnati (1990b), Lightning stroke simulation by means of the leader progression model. II. Exposure and shielding failure evaluation of overhead lines with assessment of application graphs, IEEE Trans. Power Delivery, 5(4), , doi: / Golde, R. H. (1973), Lightning Protection, Edward Arnold, London. Hussein, A. M., M. Milewski, W. Janischewskyj, F. Noor, and F. Jabbar (2007), Characteristics of lightning flashes striking the CN Tower below its tip, J. Electrost., 65(5 6), , doi: /j.elstat Krider, E. P., and S. P. Alejandro (1983), Lightning, an unusual case study, Weatherwise, 36, 71 75, doi: / Krider, E. P., and S. B. Ladd (1975), Upward streamers in lightning discharges to mountainous terrain, Weather, 30, 77 81, doi: / j tb05282.x. Les Renardieres Group (1977), Positive discharges in long air gaps at Les Renardieres 1975 results and conclusions, Electra, 53, Les Renardieres Group (1981), Negative discharges in long air gaps at Les Renardieres 1978 results, Electra, 74, Lu, G., and W. Walden-Newman (2009), Analysis of the video recordings of an aborted upward leader at daytime from a Bolt-from-the-Blue discharge, J. Lightning Res., 1, 22 27, doi: / Lu, W., Y. Zhang, L. Chen, E. Zhou, D. Zheng, Y. Zhang, and D. Wang (2010), Attachment processes of two natural downward lightning flashes striking on high structures, paper presented at 30th International Conference on Lightning Protection, Power and Energy Soc., Cagliari, Italy. Lu, W., L. Chen, Y. Zhang, Y. Ma, Q. Yin, S. Chen, and Y. Zhang (2011), Unconnected upward leaders observed in Guangzhou during , paper presented at 7th Asia-Pacific International Conference on Lightning, Tsinghua Univ., Chengdu, China. Mazur, V., L. H. Ruhnke, A. Bondiou-Clergerie, and P. Lalande (2000), Computer simulation of a downward negative stepped leader and its interaction with a ground structure, J. Geophys. Res., 105(D17), 22,361 22,369, doi: /2000jd Orville, R. E. (1968), Photograph of a close lightning flash, Science, 162, , doi: /science Rakov, V. A., and M. A. Uman (2003), Lightning: Physics and Effects, Cambridge Univ. Press, Cambridge, U. K. Rizk, F. A. M. (1994a), Modeling of lightning incidence to tall structures. I. Theory, IEEE Trans. Power Delivery, 9(1), , doi: / Rizk, F. A. M. (1994b), Modeling of lightning incidence to tall structures. II. Application, IEEE Trans. Power Delivery, 9(1), , doi: / Wang, D., N. Takagi, T. Watanabe, H. Sakurano, and M. Hashimoto (2008), Observed characteristics of upward leaders that are initiated from 11 of 12

12 a windmill and its lightning protection tower, Geophys. Res. Lett., 35, L02803, doi: /2007gl Wang, D., T. Watanabe, and N. Takagi (2011), A high speed optical imaging system for studying lightning attachment process, paper presented at 7th Asia-Pacific International Conference on Lightning, Tsinghua Univ., Chengdu, China. Warner, T. A. (2010), Upward Leader development from tall towers in response to downward stepped leaders, paper presented at 30th International Conference on Lightning Protection, Power and Energy Soc., Cagliari, Italy. Zhang, H., D. Wang, W. Lu, Q. Meng, and Y. Zhang (2012), A single-stationbased 3D lightning channel imaging system using differential arrival time of thunder (in Chinese), Plateau Meteorol., 31(1), of 12