Corresponding author: B. Zhu, CAS Key Laboratory of Geospace Environment, School of Earth and Space Sciences, University of Science and Technology of China, Hefei, China. (Zhuby@ustc.edu.cn)
 Using a three-station wideband electric and magnetic field measuring network in conjunction with VHF receivers, we recorded the so-called compact intracloud discharges (CIDs) in the Greater Khingan Range (51°N) in Northeast China. This type of lightning discharge had never been observed in such high-latitude regions before. During the summer seasons of 2009 and 2010, a total of 493 CIDs with positive electric field pulses (negative charge moving up) occurring in 31 thunderstorms was recorded, with 87% (27/31) of CID-producing storms generating less than 20 CIDs. The CID occurrence observed at 51°N appears to differ significantly from that in most lower latitude regions. Specifically, CIDs accounted only for 0.034% of the total of 1.4 million recorded lightning events, and no CIDs with negative electric field pulses (positive charge moving up) were observed. A total of 202 CIDs was located and they clustered at altitudes ranging from 5 to 12 km above ground level (agl), with a median height of 7.9 km agl. The effective ionosphere heights ranged from 75 to 95 km during the time period of 15:00–06:00 (local time). The median maximum heights of the 30 dBZ radar echo for convective cells with CIDs and without CIDs were 10.8 and 9.4 km, respectively. We infer that a relatively small vertical extent of thunderclouds, limited by the height of the tropopause, and a lack of vigorous convective surges in higher-latitude regions might be responsible for the paucity or absence of CIDs in general and apparent suppression of those with negative electric field pulses in particular.
 The distinct intracloud lightning discharges characterized by (1) extremely strong high frequency and very high frequency (HF/VHF) radiation and (2) large-amplitude and short-duration very low frequency and low frequency (VLF/LF) bipolar electric field pulses [Le Vine, 1980], have been referred to as compact intracloud discharges (CIDs), because of their short estimated channel lengths [e.g., Smith et al., 1999; Nag and Rakov, 2010; Liu et al., 2012], or narrow bipolar events (NBEs) [e.g., Eack, 2004; Jacobson and Heavner, 2005; Wiens et al., 2008]. During the past two decades or so, these discharges have attracted a great deal of attention due to their unusual electromagnetic signatures [Willett et al., 1989; Smith et al., 1999; Nag et al., 2010], their potential relationship to thunderstorm convection strength [Wiens et al., 2008], and their still unclear physical mechanism [Gurevich and Zybin, 2005; Nag and Rakov, 2010]. Ground-based observations of CIDs were first reported by Le Vine , who worked in Florida and Virginia, and then by Willett et al. , who worked in Florida. Subsequently, there were many reports on this phenomenon from different regions of the world, including New Mexico, Texas, Florida, Oklahoma, Kansas, Colorado, and Nebraska in the United States [Smith et al., 1999, 2002, 2004; Eack, 2004; Suszcynsky and Heavner, 2003; Jacobson and Heavner, 2005; Wiens et al., 2008; Nag et al., 2010, and others]; Sri Lanka [Sharma et al., 2008]; Malaysia [Ahmad et al., 2010]; Osaka in Japan [Wu et al., 2013]; and Shanghai, Guangzhou, Chongqing, and Hengdian in China [Zhu et al., 2010; Wu et al., 2011, 2012; Liu et al., 2012; Wang et al., 2012]. These observations of CIDs were made at relatively low latitudes (<42°N), while no CIDs were observed in Uppsala, Sweden (about 60°N) with the same instrumentation as that used in Sri Lanka, where CIDs were an appreciable fraction of the overall lightning activity [Sharma et al., 2008]. In this study, we make another attempt to observe CIDs in a relatively high latitude region, in Jiagedaqi, China, located in the Greater Khingan Range area, using the same instrumentation as that previously employed in Shanghai [Zhu et al., 2010]. The location of Jiagedaqi (51°N), as well as the locations of four other sites where CID observations were previously performed in China, is shown in Figure 1.
2 Instrumentation and Methodology
 During the summer thunderstorm seasons of 2009 and 2010, three stations named DaZY, KeYZ, and XinT, separated by about 100 km from each other, were deployed in Jiagedaqi (51°N), at the Greater Khingan Range, Northeast China. Each station was equipped with a VHF receiver with a bandwidth of 112.5–117.5 MHz and a VLF/LF receiver, which consisted of a vertical electric field antenna and two orthogonal magnetic field loop antennas, with a bandwidth of 800 Hz to 400 kHz. The vertical electric field, two orthogonal horizontal magnetic field components, and the envelope of VHF radiation from lightning discharges were recorded simultaneously using a four-channel digitizer at a sampling interval of 0.64 µs and a record length of 5.2 ms, with the dead time between two consecutive records being less than 30 ms. From field experience, our system could effectively record lightning discharges within 300 km, and even some strong lightning VLF/LF signals generated by CIDs and strongest return strokes at distances as large as 500 km. The VHF receiver had a low gain and was mainly used to help recognize CIDs which are expected to produce the strongest VHF emission in lightning records. CIDs were identified based on VLF/LF criteria similar to those used by Smith et al.  (also by Suszcynsky and Heavner  and Wiens et al. ), plus the occurrence of the strongest VHF emission. The events with (1) relatively short total duration (~30 µs), (2) relatively short initial half-cycle (~10 µs) of VLF/LF bipolar pulse, (3) post-trigger signal-to-noise ratio (SNR) value [Smith et al., 2002] greater than 23 dB, and (4) VHF emission three to five times greater than that of cloud-to-ground discharges or intracloud discharges (if detectable) are classified as CIDs. The other events that triggered the system are classified as “normal” lightning events. The same CID identification criteria were employed in the Shanghai study [Zhu et al., 2010]. The VLF/LF pulse (electric and magnetic field components), which is referred to as narrow bipolar pulse or NBP, and the envelope of VHF radiation of a typical CID are shown in Figure 2. The physics sign convention according to which a downward directed electric field vector is considered to be negative [e.g., Rakov and Uman, 2003] is used throughout the paper.
 The digitizer was triggered on the sum of the magnitudes of two orthogonal magnetic field components, so that triggering was independent of the polarity of NBP and, hence, on the direction of charge transfer at the source. Time stamping was achieved via a GPS receiver with a timing accuracy of about 40 ns, and two-dimensional (2-D) CID locations were obtained by using the time of arrival (TOA) technique. The source location accuracy depends on the time-measuring error at each site, as well as on the source location with respect to the observation network. The possible timing error includes the sampling time error (sampling interval of 0.64 µs in our system), the GPS timing error (about 40 ns in our system), and the timing error introduced by propagation effects on magnetic field waveforms (assumed to be about 1 µs for every 100 km). Taking into account all these timing errors, the 2-D location accuracy of our system was numerically simulated. The results are presented in Figure 3. As seen in Figure 3, the location error ranges from about ±250 m inside the network to ±10 km outside the network, depending on the source location with respect to the network. Once the 2-D location of the CID was obtained based on the TOA method, the CID heights (HCID) and virtual ionosphere heights (HI) were estimated using the differences in the arrival times of the ground wave pulse and the ionosphere-reflected and ground-ionosphere-reflected pulses [Smith et al., 1999; Wu et al., 2012]. In the case when multiple (two or three) stations detected reflected pulses, the CID height was calculated as the mean of heights based on the data from individual stations. Note that in this study CID heights as well as virtual ionosphere heights were derived only for those regions where the simulated location error was less than 10 km. The height estimation uncertainties for these CIDs were less than 1 km.
3 Data and Results
 During the field campaign in 2009 and 2010, more than 40 thunderstorms occurred around the observation site, of which only 31 thunderstorms were observed to produce CIDs. A total of 493 positive CIDs was recorded at KeYZ, which accounted only for 0.034% of the total of 1.4 million lightning events recorded at that station. The VLF/LF pulses of CIDs had the mean total duration of 27 µs and mean initial half-cycle width of 7.8 µs. Table 1 summarizes the data acquired during the 2 year observation period. In 2009, a total of 204 CIDs was detected during 11 storms, with the largest number (96) of CIDs being observed in the storm that occurred on 5 August 2009. In 2010, a total of 289 CIDs was detected during 20 storms, with the largest number (103) of CIDs being observed in the storm that occurred on 11 July 2010. The latter storm was examined in the case study by Lü et al. . Note that, except for the two storms in 2009 and two storms in 2010, which produced more than 50 CIDs each, the majority (87%) of CID-producing storms generated less than 20 CIDs each (only several CIDs in some storms). Thus, the CID activity over this region seems to be much weaker compared to the storms in lower latitude regions (see Table 2). Further, it is interesting to note that the initial polarity of all CID electric field pulses recorded during this study is positive (negative charge moving up), while both polarities were observed in most lower latitude regions [e.g., Smith et al., 2004; Wiens et al., 2008; Zhu et al., 2010; Wu et al., 2011; Nag et al., 2010]. Although our system had a dead time (less than 30 ms), we consider the probability of missing a CID to be very low, particularly in view of uninterrupted observations conducted at three independent stations during the two summer seasons. CIDs with initially negative electric field pulses (positive charge moving up), which are thought to occur between the upper positive charge region and the negative screening charge layer near the upper cloud boundary [Smith et al., 2004; Wu et al., 2012], were apparently suppressed during the higher-latitude storms examined in this study.
Table 1. Summary of CID Data Obtained During Summer Seasons of 2009 and 2010
Number of CIDs per storm
Number of CID-producing storms
Table 2. Occurrence of CIDs With Positive (Negative Charge Moving Up) and Negative (Positive Charge Moving Up) Electric Field Pulses (NBPs)
Percentage of CIDs
CID Sample Size
Percentage of +NBPs
Percentage of −NBPs
New Mexico, Texas, Florida and Nebraska (~25°N–42°N), U.S.
 Figure 4 shows the diurnal distribution of normal lightning events and CIDs during the summer seasons of 2009 and 2010. The distributions are plotted using 30 min intervals. As seen in Figure 4, the diurnal variations of normal lightning discharges show a unimodal distribution with a peak around 15:00 (local time, LT) and a broad minimum from 02:00 to 12:00 (LT), which is typical of the diurnal distribution of lightning activity over land [Zipser et al., 2006]. However, CID activity does not show peaks similar to those seen for normal lightning discharges, and the CID activity peak (although not pronounced) tends to lag the normal lightning activity peak by hours. Note that the abrupt increases in the number of CIDs during 02:30–03:00 (LT) in 2009 and during 05:30–06:00 (LT) in 2010 are mainly due to two thunderstorms that were prolific CID producers. Although CIDs in this study were collected over a period of 2 years, sample sizes for CIDs are rather small. More observations are needed to obtain a more reliable diurnal distribution for CIDs.
 In some particular storms, CIDs clustered during the active period of normal lightning activity, as observed for two thunderstorms in 2010 by Lü et al. [2012, see their Figure 8], which is consistent with the results reported by Smith et al. . A similar trend was observed for two prolific CID producers in 2009. However, the lack of CIDs during normal lightning activity peak suggests that thunderstorms with high normal lightning activity produce a relatively small number of CIDs. Recall that 87% of CID-producing storms generated less than 20 CIDs. Previous studies conducted in lower latitude regions have shown that CID activity is correlated with the strength of thunderstorm convection or normal lightning activity during the CID-producing periods [Jacobson and Heavner, 2005, Suszcynsky and Heavner, 2003; Wiens et al., 2008], although some of the strongest convections with high normal lightning rates do not produce CIDs [Wiens et al., 2008]. It appears that previous observations, as well as those shown in Figure 4, are indicative of CID activity being related to some particular thunderstorms or some special circumstances during thunderstorm development. As of today, specific conditions that lead to production of CIDs either in the lower or higher-latitude regions remain unclear.
 Figure 5 shows CID heights (HCID) and virtual ionosphere heights (HI) versus local time and their histograms for 202 CIDs that occurred during 15:00–06:00 (LT) and showed detectable ionosphere-reflected and ground-ionosphere-reflected pulses following the ground wave pulse. Note that heights of CIDs and the ionosphere given here are above ground level (agl), with the altitude of terrain around the observation site being about 500 m above mean sea level (amsl). As seen in Figure 5, HI ranged from 75 to 95 km, which is consistent with the results of Smith et al.  and Han and Cummer [2010a, 2010b]. Most of the CID heights are in the range of 5–12 km agl, with a geometric mean height of 8.1 km agl and a median height of 7.9 km agl. Our mean height of CIDs is somewhat lower than those reported in lower latitude regions [e.g., Smith et al., 2004; Zhu et al., 2010; Wu et al., 2012; Wang et al., 2012; Nag et al., 2010].
 Figure 6 shows the locations of normal lightning events and CIDs during 03:16–03:26 (LT) superimposed on the PPI (plan position indicator) radar reflectivity image at 03:23 (LT) on 8 August 2009, which is during the active period, as seen from the radar reflectivity data throughout the thunderstorm lifetime. Radar reflectivity was obtained from an operating meteorological CINRAD/CC radar located about 10 km north of KeYZ station and at the origin of coordinate system in Figure 3. The radar performed volume scans at nine elevation angles (0.5°, 1.5°, 2.4°, 3.4°, 4.3°, 6.0°, 9.9°, 14.6°, and 19.5°) every 6 min, with sweep radius of 150 km and spatial resolution of 600 m. The reflectivity data at the elevation angle 2.4° were used in Figure 6. Lightning events, including CIDs, during this time period occurred in regions where our observation network has the best location accuracy (see Figure 3). As seen in Figure 6, normal lightning discharges scatter over several convective regions, while CIDs cluster only in two regions with high radar reflectivity. A total of 11 CIDs is located at heights ranging from 6.9 to 9.4 km agl (7.4–9.9 km amsl), with a median height of 8 km agl (8.5 km amsl). From the vertical profile of the radar reflectivity along the line AB, most of CIDs cluster in close proximity to the convection cell within the 35 dBZ contour of radar reflectivity and at the upper boundary of the most intense convection regions. The maximum height of the 30 dBZ radar reflectivity for CID-producing convective cells is limited to about 10–12 km. Similar results from another prolific CID-producing thunderstorm in 2010 were discussed by Lü et al. .
 The PPI image in Figure 6 shows that CIDs apparently tend to selectively occur in certain convective cells. In order to check the differences between convective cells with CIDs (CID-producing cells) and those without CIDs, we compare the maximum heights of the 30 dBZ (H30dBZ-max) radar echo for all CID-producing and non-CID-producing convective cells during the two summer seasons, as illustrated in Figure 7. A total of 251 individual convective cells within the scope of radar that produced detectable lightning events was found, among which only 74 produced CIDs. The H30dBZ-max for each individual convective cell was obtained from the vertical profile of the convective core.
 As seen in Figure 7, normal lightning discharges occurred in convective cells with H30dBZ-max as low as about 6 km, while CIDs occurred only in convective cells with the lowest H30dBZ-max of about 8 km. Further, the most frequent range of H30dBZ-max for CID-producing cells is about 9–12 km, which is consistent with the illustration for the case study presented in Figure 6 and work of Lü et al. . The higher values of H30dBZ-max (the lowest value, the median value, and the highest value) for CID-producing cells imply that, in general, CIDs tend to occur in stronger convective cells as represented by H30dBZ-max. However, Figure 7 shows that a lot of strong convective cells with H30dBZ-max as high as 12–13 km did not produce CIDs, which is similar to the findings reported by Wiens et al.  and suggests that CIDs require still unknown conditions for their occurrence.
 The maximum height of the 30 dBZ (H30dBZ-max) radar echo for CID-producing cells as well as non-CID-producing cells in the higher-latitude region is appreciably lower than that for CID-producing storms in lower latitude regions, such as the U.S. Great Plains, where the maximum height of 30 dBZ reflectivity in different thunderstorms was usually greater than 14 km [Wiens et al., 2008], and in Guangzhou where the maximum height of 30 dBZ reflectivity was greater than 15 km [Wu et al., 2012]. The relatively low height of the 30 dBZ (H30dBZ-max) radar echo for CID-producing cells in the higher-latitude region is indicative of relatively weak convection in that region. It may be influenced by the relatively low height of the tropopause in such regions. From the global-merged infrared cloud-top brightness temperature imagery (Meteosat-7 satellite, 11.5 µm wavelength), all the CIDs examined here were located within the 225K (−48°C) contour, with a minimum cloud top temperature of 215K (−58 °C), which is somewhat warmer than the cloud top temperature of CID-producing thunderstorms (usually −50°C to −60°C) in lower latitude regions [Jacobson and Heavner, 2005].
 As mentioned in section 1, characteristics of CIDs observed in the relatively low latitude regions (<42°N) have been documented by several research groups. Tables 2 and 3 summarize those characteristics.
Table 3. Median Heights (agl) of CIDs With Positive and Negative Electric Field Pulses (NBPs) and Ranges of Height Variation (in the Parentheses)
 Table 2 gives the occurrence of CIDs with positive and negative initial polarity electric field pulses. Wiens et al.  reported the occurrence of a total of 34,046 CIDs out of over eight million lightning events recorded by LASA (Los Alamos Sferic Array) during May to July in 2005 in the U.S. Great Plains. CIDs with positive electric field pulses (+NBPs) accounted for about 77% of all CIDs. Smith et al.  reported that in over seven million lightning discharges recorded by LASA during more than 4 years of observation, more than 100,000 CIDs were identified, with the percentage of +NBPs being approximately 58%. In Florida, Jacobson and Heavner  reported that a total of 103,240 CIDs was recorded during 1999–2002, and 77% of those produced +NBPs, while during 2001–2002, CIDs with +NBPs accounted for 63% of the events [Suszcynsky and Heavner, 2003]. Interestingly, in Florida, CIDs with +NBPs and −NBPs tended to occur in different storms [Jacobson and Heavner, 2005], while in the Great Plains, they often occurred in the same storms [Wiens et al., 2008], although Wu et al.  questioned this disparity and stated that in Guangzhou, China, all nine examined storms produced CIDs of both polarities. Wu et al.  reported unusually frequent occurrence of CIDs in two locations in China. Specifically, in Guangzhou (South China), during 19 days in 2007, 7882 CIDs with +NBPs (66%) and 3994 CIDs with −NBPs were detected, and in Chongqing (Southwest China), during about 2 months in the summer of 2010, 36,442 CIDs with +NBPs (82%) and 7893 CIDs with −NBPs were recorded. In the case study of CIDs during 3 h of a thunderstorm in Shanghai, Zhu et al.  reported that CIDs with +NBPs accounted for 87% of 77 CIDs. In Osaka (~35°N), a total of 232 CIDs with +NBPs and 22 CIDs with −NBPs was observed during the summer of 2012 [Wu et al., 2013]. However, Wang et al.  recorded 236 CIDs with only +NBPs during 6 h of a storm in Hengdian (~38°N), and Sharma et al.  observed 21 CIDs in Colombo, Sri Lanka, all of which had +NBPs. In contrast to most of the lower latitude observations, in our study at the Greater Khingan Range, CIDs were less numerous and their corresponding NBPs came only in one polarity. More details on our findings are given below.
 Only 493 CIDs were recorded during the 2 year observation period at the Greater Khingan Range. CIDs accounted for as little as 0.034% of all lightning events, which was significantly less than that in lower latitude regions. Further, 27 out of the 31 CID-producing storms generated less than 20 CIDs, although Smith et al.  recorded as few as 5, 11, and 8 CIDs in the three storms in New Mexico and West Texas that they studied.
 No CIDs with −NBPs (positive charge moving up) were detected in the Greater Khingan Range area during 2 years of observations. In contrast, CIDs with both +NBPs and −NBPs were recorded in most lower latitude regions, although CIDs with +NBPs tended to occur more frequently [Smith et al., 2002; Jacobson and Heavner, 2005; Wiens et al., 2008; Wu et al., 2012, 2013; Nag et al., 2010]. Note that our recording system was capable of recording both positive and negative signals. Thus, we interpret the results of this study as indicative of the rare occurrence (if at all) of CIDs with −NBPs in higher-latitude regions.
 Table 3 gives the median heights of CIDs with positive and negative electric field pulses (NBPs). As seen from this table, in lower latitude regions, CIDs with +NBPs (negative charge moving up) and −NBPs (positive charge moving up) tend to occur at different altitudes. The median height for +NBPs ranged from 9.9 to 15 km in the lower latitude regions, and for −NBPs it clustered at 15–17 km. Although the source heights for CIDs with +NBPs had a large range of variation in different regions, their distributions peaked at about 9–10 km agl in Shanghai [Zhu et al., 2010] and Chongqing [Wu et al., 2012], 11–12 km agl in Guangzhou [Wu et al., 2012] and Hengdian [Wang et al., 2012], 13–15 km agl in Florida [Smith et al., 2004; Nag et al., 2010], and ~5–15.5 km in Osaka [Wu et al., 2013]. CIDs with +NBPs in the Greater Khingan region clustered at 5–12 km agl, with a median height of 7.9 km agl, which was slightly lower than the geometric mean height of CIDs with +NBPs in Shanghai [Zhu et al., 2010] and Chongqing [Wu et al., 2012] and much lower than the median or geometric mean height in other lower latitude regions, such as Florida [Smith et al., 2004; Nag et al., 2010] and Guangzhou [Wu et al., 2012].
 Finally, we discuss possible explanations of the fact that there were no CIDs with −NBPs recorded in the Greater Khingan Range, in contrast with observations in lower latitude regions [Smith et al., 2004; Nag et al., 2010; Wu et al., 2012, 2013]. One possible explanation is a smaller vertical extent of higher-latitude clouds. The relatively weak convection in thunderstorms developing in higher-latitude regions is expected and can be seen from the relatively low maximum height of 30 dBZ radar echo, as illustrated in Figures 6 and 7. Wu et al.  compared CID heights observed in Chongqing and Guangzhou and put forward the hypothesis that there was a “critical height” for the production of CIDs with +NBPs (about 7 km) and −NBPs (about 15 km). From this point of view, the relatively weak convection (typical maximum height of the 30 dBZ radar reflectivity of 10–12 km, with 74% of all the convective cells having the H30dBZ-max in the range from 9 to 12 km) may be not conducive to the creation of a favorable electrical environment above the main (upper) positive charge region for the occurrence of CIDs. It is known that CIDs are often associated with strong convection [Jacobson and Heavner, 2005; Wiens et al., 2008] or the period of rapid intensification of a hurricane [Fierro et al., 2011], and CID heights generally increase with increasing CID flash rates [Suszcynsky and Heavner, 2003]. With regard to CIDs with −NBPs, they tend to occur at higher altitudes, and the percentage of −CID increases with increasing convective strength of a thunderstorm [Wu et al., 2011, 2013]. The relatively weak convection, limited by the lower height of the tropopause in higher-latitude regions, may be responsible for both the relatively infrequent occurrence of CIDs with +NBPs and their slightly lower source heights, as well as for the absence of CIDs with −NBPs there. It appears that conditions for initiation of CIDs with −NBPs occur less frequently than those for initiation of CIDs with +NBPs, and the former are more easily suppressed when the cloud top height is relatively low. Note that in the highest-latitude region (~60°N) in Sweden, where a search for CIDs was performed, neither negative nor positive NBPs were detected [Sharma et al., 2008]. Based on the above facts and considerations, we suggest that the differences in the vertical extent of thunderstorm convection, which is related to the height of the tropopause above which convection is suppressed, could explain the observed differences in CID production in the Great Khingan Range area relative to the lower latitude regions.
 However, the results presented here, as well as those reported by other authors [e.g., Wiens et al., 2008; Wu et al., 2012], suggest that conditions inside thunderstorms favoring the production of CIDs are rather complex. On the one hand, CIDs tend to occur in thunderstorms with severe convection, which are characterized by intense normal lightning activity [Smith et al., 1999; Suszcynsky and Heavner, 2003], the lower cloud top temperature [Jacobson and Heavner, 2005], and the higher maximum height of the 30 dBZ radar echo, as found in this study and by Wiens et al. . It appears that CIDs can only occur in certain electrical conditions which are determined to a large extent by convection strength, particularly in the case of CIDs with −NBPs as illustrated by Wu et al. [2012, 2013]. On the other hand, CIDs tend to selectively occur in some particular convective cells; that is, many strong convective cells do not produce detectable CIDs, as shown in this study and by Wiens et al. . Thus, it appears that strong convection is a necessary, but not sufficient condition for CID occurrence. It is likely that the detailed electrical structure of the upper part of the cloud (including convective surges overshooting the tropopause) and “external” effects, such as runaway breakdown initiated by an energetic cosmic ray particle [Gurevich and Zybin, 2005], each play an important role. Unfortunately, neither our study nor previous ones provide sufficient clues on how exactly CIDs are initiated. Without such clues, any discussion of CID occurrence is necessarily speculative. Clearly, further research is needed.
 Observations of CIDs in the relatively high-latitude region (51°N) in Northeast China are presented. A total of 493 CIDs that occurred in 31 thunderstorms was recorded during the 2 year period, and 87% (27 out of 31) of those thunderstorms produced less than 20 CIDs. The occurrence of CIDs relative to the total number of recorded lightning events was much lower compared to that in lower latitude regions. All the recorded CIDs were associated only with +NBPs and were located at altitudes between 5 and 12 km, with a median height of 7.9 km agl. The median maximum heights of the 30 dBZ radar reflectivity for thunderstorm cells with CIDs and without CIDs were 10.8 and 9.4 km, respectively. We infer that the relatively low cloud top heights in the higher-latitude region might be responsible for both the infrequent occurrence of CIDs in general and the absence of CIDs with −NBPs (positive charge moving up) in particular.
 This study was supported in part by the R&D Special Fund for Public Welfare Industry (Meteorology) under grant GYHY201006005, the Fundamental Research Funds for the Central Universities WK2080000031, and the National Natural Science Foundation of China under grant 41075001. It was also supported in part by the U.S. National Science Foundation grant ATM-0852869 and DARPA grant HR0011-10-1-0061.