We have analyzed the electric field changes of 14 upward leaders that were initiated from a windmill and its lightning protection tower. We found that these upward leaders can be sub-classified into two types according to whether they are triggered by nearby lightning discharge activity or they are initiated without any nearby preceding discharge activities. We have also obtained evidence of upward aborted leaders that are initiated from high-grounded objects in response to nearby lightning. All these results suggest that once the electric field which surrounds the high-grounded object is large enough, an upward leader can be initiated no matter whether the electric field is built up slowly or created rapidly by a nearby lightning discharge, although the later appears to be more efficient in triggering the leader. In addition, we found that without the assistance of a rising electric field produced by a nearby lightning discharge, compared to a stationary windmill and a tower with a similar height, a rotating windmill tends to have a bigger chance of initiating an upward leader.
 In Japanese winter thunderstorms, many cloud-to-ground lightning discharges are characterized as upward lightning [e.g., Miyake et al., 1990; Suzuki, 1992]. These upward lightning discharges have caused a lot of damage to high objects, such as large-scale windmills. A comprehensive review of upward lightning as well as of Japanese winter lightning has been given by Rakov and Uman [2003a, 2003b]. As is generally known, an upward lightning is characterized by an upward leader that is initiated from a high-grounded object and that propagates toward overhead charged clouds. An interesting question about upward lightning is how its upward leader is initiated. Considering that a slower charge build-up allows time for corona discharge to inhibit the initiation of an upward leader, Berger and Vogelsanger  suggested that the high electric field needed for the initiation is rapidly created by an in-cloud discharge. In other words, an upward leader is assumed to be initiated by preceding in-cloud discharge activity. Based on simultaneous observation of E-fields and high-speed images, Takagi et al.  claimed that all nine upward positive leaders observed by them were initiated without any apparent in-cloud discharge activity. In this paper, we present evidences showing that 10 of 14 upward leaders analyzed are initiated by nearby lightning discharge activity while the remaining 4 are initiated without any apparent preceding discharge activities.
2. Observation and Data Description
 As shown in Figure 1, a total of 5 observation sites are setup around a windmill and its lightning protection tower, which is located at Uchinada-chou, Ishikawa prefecture of Japan, with their heights being 100 m and 105 m, respectively. The windmill and the tower are separated at a distance of about 45 m, and are built on a small hill just adjoining the sea, the height of the hill being 40 m above sea level. At 3 of the 5 observation sites, Center, Chitori and Hori, both a slow antenna [e.g., Brook et al., 1982] and a fast antenna [e.g., Weidman and Krider, 1978] for measuring electric field changes of lightning discharges were operated, while at the remaining two sites, Ooba and Shirao, only a fast antenna was available. The slow antenna has a frequency bandwidth from several Hz to around 3 MHz. The fast antenna has a frequency bandwidth from several kHz to around 5 MHz. The fast antenna is about 100 times more sensitive than the slow antenna. The outputs of all antennas are recorded with digital oscilloscopes which are operated at a sampling rate of 10 MHz and a recording time of 0.8 s for each triggered event. All the recording systems are GPS synchronized with a time accuracy of about 100 ns. Additionally, at the Center site, various optical observation systems, including two general video cameras and a high-speed video digital camera (Kodak, HS-4540) were used for the present study.
 During our observation period from December of 2005 to February of 2006, we recorded a total of 21 lightning flashes which struck on the windmill or/and its lightning protection tower as shown in Table 1. From video recordings, 8 of them are identified striking on the windmill, 12 on the tower and the remaining one on both the windmill and the tower. For these lightning, only part of the corresponding electric field change data are obtained due to abnormally heavy snow at our observation sites. To avoid confusing, in this study we present the results of only the 14 lightning flashes whose electric field changes have been recorded at Center site.
Table 1. List of Lightning That Struck a Windmill and/or Its Lightning Protection Tower and Were Recorded With Our Video Camera Systems From December of 2005 to February of 2006a
Lightning Striking Location
Leader Direction and Polarity
State of Windmill
The 14 of 21 lightning that were analyzed in the present study.
 A detailed comparison of the electric field changes recorded at Center site has been made and it is found that the electric field changes can be apparently divided into two types as shown in Figure 2. The first type, as shown in Figure 2a, exhibits suddenly a rapidly-decreasing electric field change which leads to the following saturation. The second type, as shown in Figure 2b, exhibits at least two changes which are in opposite directions. For convenience, these two types of changes are referred as type 1 and type 2 changes. The corresponding leaders and lightning are called as type 1 and type 2 leaders and lightning, respectively.
 To understand the difference between these two types of lightning, the electric field changes simultaneously recorded at three sites are compared. Figure 3a shows an example of type 1 electric field changes which were recorded at sites of Center, Chitori and Horii at 0:36:29, Jan.3rd, 2006. An upward-directed electric field is defined as positive. As seen in Figure 3a, although all the electric field changes are saturated, the following three features can be identified: (1) The electric field changes at all three sites varied in the same direction; (2) Electric field changes of return strokes are not identified immediately after the initial change; (3) The electric field changes at Center site, which is the closest site to the windmill and the tower, drops earliest and most steeply. All these features combined suggest that the initial electric field changes are produced by an upward-propagating positive leader from the windmill. Figure 3b shows an example of the type 2 electric field changes which were recorded at 21:25:21, Dec. 25th, 2005. As shown in Figure 3b, the electric field changes at all sites first increase and several tens of milliseconds later begin to decrease. During the decreasing stage, the electric field changes exhibit the similar 3 features identified for type 1 leaders. Authors assume that these 3 features indicate that an upward-propagating positive leader is initiated from the tower. In contrast, during the increasing stage, the electric field changes at the distant sites of Chitori and Hori are larger than that at Center site. An upward-propagating leader initiated from the tower or/and the windmill is not able to produce such electric field changes. These electric field changes must be produced by a lightning discharge occurring in another place, such as in-clouds. This preceding discharge increases rapidly the ambient electric field of the windmill and the tower as shown in Figure 3b, and consequently initiates the following upward-propagating leader from the tower.
 The classification of all the 14 leaders analyzed is summarized in Table 1. All these leaders are identified to be upward-propagating leaders. Among them, 4 are Type 1 leaders and the remaining 10 are type 2 leaders. In term of leader polarity, 12 are positive leaders and only 2 are negative leaders. Both negative leaders are type 2 leaders.
3.2. Evidence of Aborted Upward Leaders in Response To Nearby Natural Lightning
 During our observation campaign, we observed many aborted upward leaders which occurred from both the windmill and the tower in response to nearby lightning. Evidence of such aborted upward leaders was shown in Figure 4, which was recorded with our high-speed video camera being operated at a frame interval of 222 μs or so and covering an effective area of about 200 m × 200 m for the windmill and the tower. In response to a nearby lightning, which occurred on 23:32:01, Jan.2, 2006, totally 4 aborted upward leaders occurred. The first one, as seen in Figure 4, was initiated from windmill and lasted two frames (No. 1–2). The second one was from the tower and lasted also 2 frames (No. 5–6). The third one was again from tower and lasted 4 frames (No. 15–18). The fourth one, appearing as the brightest one, was from windmill and lasted also 4 frames (No. 22–25). The speeds of the last two leaders are estimated to be about 1 × 105 m/s, similar to that of general lightning leaders. The electric field change of the nearby lightning is shown in Figure 5. The 4 aborted leaders (AL1–4) occurred about 50 ms after the initiation of the lightning. Since the aborted leaders occurred during a negative electric field change, it is inferred that all the 4 aborted leaders are of negative upward-propagating types.
4. Discussion and Concluding Remarks
 Our observational results indicate that upward lightning can be subdivided into two types according to whether their upward leaders are triggered by a nearby lightning discharge activity or they occur without any preceding nearby discharge activities. In addition, we have obtained evidence of upward aborted leaders occurred from high-grounded objects in response to nearby lightning. These results suggest that a high-grounded object is a strong competitor for initiating a discharge leader. Once the electric field which surrounds the high-grounded object is large enough, an upward leader can be initiated no matter whether the electric field is built up slowly or created rapidly by a nearby lightning discharge, although the later appears to be more efficient in triggering the leader. It has been reported that many upward lightning in Japanese winter thunderstorm involved more than one high-grounded object [e.g., Miyake et al., 1990]. It is possible that such multiple upward discharges are initiated by the same in-cloud discharge, whose horizontal extent largely determines the number of, and the distances between, the objects involved, as pointed out by Rakov and Uman [2003a, 2003b]. It should also be possible that once an upward lightning occurred from a high-grounded object, this upward lightning triggers other upward lightning of opposite polarity from nearby high-grounded objects if favorable electrical charge structures of thunderstorms are available. An example of such favorable electrical structure is shown in Figure 6. In this figure, UL1 is assumed to be initiated from tower 1 without any preceding in-cloud activity while UL2 is triggered by UL1 from tower 2. Due to presence of high-grounded objects, additional cloud-to-ground lightning discharges are likely to occur.
 For the 14 lightning flashes analyzed in the present study, it is interesting to note that among the 5 lightning flashes on windmill, 3 are type 1 and 2 are type 2, while among the 8 lightning on tower, 1 are type 1 and 7 are type 2. Apparently, the tower has a bigger chance to be hit by type 2 lightning and the windmill by type 1 lightning. Among the 3 type 1 lightning discharges which struck the windmill, all 3 were initiated when the windmill was rotating. Moreover, on December 22 of 2005, all 4 lightning flashes hit the rotating windmill. All these facts combined suggest that, without the assistance of a rapidly increasing electric field produced by a nearby lightning discharge, compared to a stationary windmill and a tower with a similar height, a rotating windmill tends to have a bigger chance of initiating an upward leader. Thus, when a thunderstorm is overhead, in order to reduce the probability for a high windmill to be struck by lightning, it is better to stop the rotation of the windmill.
 This research was supported by Ministry of Education, Culture, Sports, Science, and Technology of Japan (Grant 17360124). The authors would like to thank K. Minamitani, T. Matsui, and D. Tsuji for their help in carrying out the experiment, Y. Hasegawa for his help in analyzing the experiment data, and M. Uman for his comments. The high-speed video camera used in the present study is borrowed from Life Science Research Center of Gifu University.