Key Laboratory of Middle Atmosphere and Global Environment Observation, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China
Corresponding author: X. Qie, Key Laboratory of Middle Atmosphere and Global Environment Observation, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China. (email@example.com)
 The current and electric field pulses associated with M-component following dart leader-return stroke sequences in negative rocket-triggered lightning flashes were analyzed in detail by using the data from Shandong Artificially Triggering Lightning Experiment, conducted from 2005 to 2010. For 63 M-components with current waveforms superimposed on the relatively steady continuing current, the geometric mean values of the peak current, duration, and charge transfer were 276 A, 1.21 ms, and 101 mC, respectively. The behaviors of the channel base current versus close electric field changes and the observation facts by different authors were carefully examined for investigation on mechanism of the M-component. A modified model based on Rakov's “two-wave” theory is proposed and confirms that the evolution of M-component through the lightning channel involves a downward wave transferring negative charge from the upper to the lower channel and an upward wave draining the charge transported by the downward wave. The upward wave serves to deplete the negative charge by the downward wave at its interface and makes the charge density of the channel beneath the interface layer to be roughly zero. Such modified concept is recognized to be reasonable by the simulated results showing a good agreement between the calculated and the measured E-field waveforms.
 The rocket-triggering lightning technique facilitates the research of lightning discharge and its effects, for the striking point of the triggered lightning is assured, which makes it possible to synchronously detect the current at the channel bottom and the associated electromagnetic field at various distances [Hubert et al., 1984; Rakov et al., 1998; Qie et al., 2007, 2011]. A negative triggered lightning generally contains a so-called initial stage (IS), characterized by an upward extending leader in positive polarity and the yielding initial continuous current [Wang et al., 1999; Willett et al., 1999; Lalande et al., 1998, 2002], and a discharge stage, characterized by one or more dart leader-return stroke sequences [Rakov et al., 2005; Schoene et al., 2009; Qie et al., 2009; Zhao et al., 2009]. In some instances, the return stroke is trailing with a continuing current on which several current pulses may be superimposed, referred to as M-components, and consequently exhibiting transitory luminosity enhancement in the already luminous channel [Malan and Collens, 1937; Fisher et al., 1993; Rakov et al., 1995, 2001; Qie et al., 2011].
 The directly measured channel base current for M-components show waveforms that typically involve a wavefront risetime of several hundred microseconds and a magnitude of several hundred amperes [Thottappillil et al., 1995], though in some particular cases, the peak current may be up to several kilo amperes [Rakov et al., 1998; Miki et al., 2002; Flache et al., 2008; Jiang et al., 2011; Qie et al., 2011]. A general M-component produces a charge transfer of approximately 100 mC from cloud to ground, which is less than that produced from a return stroke by an order of magnitude [Rakov and Uman, 2003].
Shao et al.  analyzed the observational facts of cloud-to-ground lightning in detail by using the narrow band very high frequency (VHF) interferometer system. They suggested that two types of in-cloud discharge processes (or streamer developments) may result in the occurrence of M-component. The first is characterized by a negative streamer propagating forward at speeds of 106–107 m/s and connecting into the conducting channel that flows with continuing currents. The other process is characterized by the occurrence of a positive streamer extending outward from the upper extremity of the channel and a negative streamer recoiling back with a faster speed into the channel to ground. By analyzing the high-speed camera data on those object-initiated lightning (negative polarity) from tall towers, Mazur and Ruhnke  found discharge processes similar to the first type by Shao et al. , and they interpreted that the M-component is caused by recoil leaders intercepting the conducting channel to ground. Recently, Yoshida et al.  found that M-component type processes can be initiated at the upper channel extremity by either recoil leaders or via the interception of separate in-cloud stepped leaders by a grounded current-carrying channel.
 On the basis of the synchronous current at the discharge channel bottom and the associated surface electric field measured at 30 m away, Rakov et al.  carried out a detailed analysis on the physical process of M-component in the lightning channel and proposed a so-called “two-wave” (or guided wave) mechanism. They stated that M-component would involve less sensitive decreasing of electric field magnitude with distance, as compared to leader-return stroke. The simulated results based on such mechanism were generally consistent with the multistation electric and magnetic field measurements [Rakov et al., 1995, 2001]. Zhang et al.  conducted a simulation of electromagnetic fields by using this two-wave mechanism and confirmed such a theory to be useful for explaining the M-component. Jiang et al.  found that the two-wave model-predicted E-field waveform appeared an unexpected overshooting on the recovery (trailing) edge.
 To date, many observation facts have been observed on M-component; however, there are still some controversies (regarding its behaviors or charge transferring mode) that need to be resolved or clarified. In this paper, we conducted researches on the statistical characteristics of M-components in lightning flashes triggered in the Shandong province of China from 2005 to 2010. The physical mechanisms of M-component proposed by different researchers and some existing observational facts that are valuable for investigating the evolution of M-component are reviewed and discussed in detail. A modified mechanism (based on Rakov's two-wave theory) concerning the evolution and charge transferring process of M-component through the channel is proposed.
2 Experiment and Data
 The Shandong Artificially Triggering Lightning Experiment (SHATLE) has been carried out continuously since 2005 [Qie et al., 2007; Zhang et al., 2009]. Eight rockets trailing grounded wire were installed at the launching site and fired by a set of fiber ignition system in appropriate conditions during thunderstorm to trigger lightning discharges [Qie et al., 2010]. Integrated synchronous observations were conducted for the triggered lightning flashes, including direct measurement of discharge current at the channel bottom, surface electric field and magnetic field changes with high-temporal resolution at different distances from the channel, optical evolution of the channel by high-speed camera, VHF radiation source detection, and such.
 The currents were measured by a Pearson coil with a frequency range from 0.9 Hz to 1.5 MHz and a current viewing shunt having a constant resistance of 5 mΩ within frequency coverage of 0–3.2 MHz. Both instruments were installed in a faraday cage at the lightning channel base. The current signals were transmitted to and recorded by a DL750 digital oscilloscope placed in the control room, which was 70 m away, with a sampling rate of 10 MSample/s [Qie et al., 2011]. Fast and slow flat plate antennas with time constants (decay to 1/e) of 1 ms and 3 s and frequency responses up to 5 and 2 MHz, respectively, were installed for detecting fine electric field changes associated with the development of the triggered lightning. The antennas were installed at different distances from the rocket launcher, near site was 60 m away, and far site was 450–1000 m away (there were a few variations in different years) [Qie et al., 2009; Yang et al., 2010]. Field mills were installed for continuous monitoring of the surface electric field evolution and the perpendicular crossing rectangular loops for magnetic field detection. The signals obtained from different instruments were synchronized by GPS with a time precision of ±50 ns. More information about the installation setup and the associated technical parameters can be found in Yang et al.  and Qie et al. .
 Twenty-two negative cloud-to-ground lightning flashes were successfully triggered during 2005–2010. The current at the channel bottom, together with electric field change for some of these flashes, was acquired. A total of 63 M-components were identified from the directly measured current waveforms.
3 Analysis and Results
3.1 The Current Waveform Parameters of M-Components
 Figure 1 shows the current waveforms of an M-component that occurred during the continuing current stage following behind the trail of a return stroke. The current level preceding the M-component was 35 A. Here a positive value of the current is defined as the positive charge being lowered from cloud to ground. As in Figure 1, the M-component demonstrated somehow V-shaped current waveforms that had a trailing edge being a little bit slower than the leading edge. Such a waveform characteristic is basically the same as that described by Rakov et al. . The associated background continuing currents were also found to be generally steady for most of the M-components, excluding a few of them that occurred immediately after the return stoke when the current had not fallen to a steady level.
 A total of 63 M-components are identified from triggered lightning flashes based on the directly measured current waveforms during SHATLE 2005–2010. Qie et al.  once analyzed the current waveform parameters of these M-components, together with the return strokes and ICC (initial continuous current) pulses. For the sake of clarity, Table 1 shows the statistics of the current waveform parameters of the M-components. The definitions of the waveform parameters in this paper are the same as those defined by Thottappillil et al.  and Qie et al. . The geometric mean (GM) values of the peak current (IP), 10%–90% risetime (t10–90), half peak width (tHPW), duration (tW), and charge transfer (Q) are 276 A, 251 µs, 242 µs, 1.21 ms, and 101 mC, respectively. As shown in Table 1, the peak current of the M-components is in the range of 23.6–7000 A, with the minimum being 2 orders of magnitude smaller than the maximum. Figure 2 further shows the histograms of the current waveform parameters. The current parameters of the M-components are distributed in a wide range. As evident in Figure 2, the parameter distributions do not significantly concentrate in the near range of the GM values listed above, and several samples have values that differ by 1 order of magnitude from the GM values.
Table 1. Statistics of Current Waveform Parameters of M-Components
 It has been reported that some larger than usual M-component pulses may have peak current of kilo amperes, and a few of them could even be comparable with return strokes [Qie et al., 2011]. We checked all 63 samples studied here and found that 11 of them (17%) involved peak currents exceeding 1 kA. Table 2 shows the comparison between the parameters (geometric mean values) of these 11 large pulses and all of the samples. These 11 samples have the GM peak current, 10%–90% risetime, half peak width, duration, and charge transfer of 2774 A, 57 µs, 59 µs, 0.65 ms, and 353 mC, respectively, indicating sharper waveforms and higher charge transfers than the typical ones.
Table 2. Comparison of the Current Parameters (GM Values) of Larger Than Usual M-Components and Those Acquired From All of the Samples
3.2 Electric Field Change of the M-Components
 The surface electric field changes produced by lightning are closely related to the discharge processes. Figure 3 shows the synchronous recordings of current at the channel bottom and the associated electric field change at 60 m away for the M-component shown in Figure 1c. The electric field waveforms shown in this paper were detected by the slow flat plate antenna, with a time constant of 3 s. Here a positive change of the electric field at ground corresponds to negative charge being lowered (or removed) to ground from cloud. As in Figure 3, it is clear that the electric field change of the M-component pulse exhibited overall waveforms of V shapes (with the trailing edge being a little bit steeper than the leading edge), and it begins its negative-going change when the current at the channel bottom is still in the continuing current level. When the current emerges out of the background level, the concurrent electric field detected at 60 m is still getting larger. The peak of the current at the channel bottom lagged behind the peak of the electric field at 60 m by tens to hundreds of microseconds. These features for the electric field of M-components are basically the same as those described by Rakov et al. . Merely, here we show the electric field waveforms measured at 60 m, while Rakov et al.  showed the counterparts at 30 m.
 As also shown in Figure 3, the electric field change (obtained by the slow antenna) of the M-components involved the trailing edge ending at a steady level. And generally, the electric field at close range from the lightning channel for M-component shows a slightly higher electric field level of the trailing edge than the leading edge but does not involve significant fluctuation or overshooting superimposed on the trailing edge.
4 Mechanism of M-Component
4.1 The Existing Observational Facts and Mechanism of M-Components
Flache et al.  reported that a negative flash (flash 4 in their paper) initiated from the Peissenberg tower (Germany, 160 m high), containing 14 IS pulses, 4 return stroke pulses, and 5 M-component pulses (classifications based solely on the current waveforms), involved four channel branches that originated in common from a channel section which was connected to the tower. The pulse events occurred separately in three of these branches, and their physical process was complicated. Those defined “fast” IS pulses and M-components were determined to be the results of dart leader-return strokes occurring in a certain branch (being not luminous prior to the occurrence of the pulses), and meanwhile, steady continuing current was flowing in another branch; those “slow” pulses occurring in already luminous channels were interpreted to be caused by M-component and the associated mode of charge transferring process. These facts are valuable and important, manifesting that it may be somewhat arbitrary to define a current pulse as an M-component by solely examining the existence of the continuing current, and additionally, the triggering mechanism of the M-component may be due to the intracloud process or the process from the cloud to the upper channel (or the channel extremity). The VHF mapping observations by Shao et al.  and by Yoshida et al.  concerned such intercloud processes, and their descriptions of M breakdown and streamer (or leader) developments are useful for the further investigation of the M mechanism through the channel, which will be discussed in the following section.
Miki et al.  once measured the electric fields at the very near distance from the triggered lightning channel by using the so-called Pockels sensors, which were installed at a fixed position being 10 cm from the strike rod (the horizontal distance between the sensor and the channel ranged from 0.1 to 1.6 m, for some channels attached to the surrounding ring). For 8 out of 10 M-components with recorded current waveforms, no significant electric field features were obtained by Pockels sensors, indicating that the very near electric fields, which should reasonably be due to the occurrence of these M-components, were smaller than 20 kV/m, which was the lower detecting limit of the Pockels sensors. Although the recorded signals for the other two samples were distinguishable from the background level, the electric field peaks (of these two M-components) were much smaller than their counterparts of the return strokes with comparable current amplitude. Figure 4, adopted from Miki et al. , shows the current measured at the channel bottom in association with the synchronous E-field at 0.1 m, for an M-component. It can be found that the M-component caused quite narrow E-field pulse at a very near distance from the discharge channel, which may be important for explaining the M mechanism, although it is not mentioned in their paper. As shown in Figure 4, the duration of the E-field pulse at 0.1 m is predominantly shorter than the corresponding current pulse.
 On the basis of the features of the correlated current at the channel bottom and the corresponding electric fields at 30 m (similar to those described in section 3.2), Rakov et al.  once proposed a so-called two-wave (or “guided wave”) mechanism to explain the M-component evolution, which was further supported by multistation observations of electromagnetic fields. The magnitudes of electric fields of M-component exhibited logarithmic distance dependence [Rakov et al., 2001]. According to such a mechanism, the M-component involves two guided waves that propagate through the channel with opposite directions and approximately equal amplitudes. When the downward developing incident wave gets to the ground, a mirroring (reflected) wave starts to propagate upward. The ground is sensed as a short circuit, with the reflectance for current at the ground being approximated to +1; meanwhile, the counterpart for charge density is being approximated to −1. The processes propagating downward and upward have approximately similar contributions to the total outflow of the charge from the lightning channel base at any instant of time.
Mazur and Ruhnke  studied the physical processes of upward lightning by analyzing high-speed video images and electric field measurements. They interpreted that the impulsive luminous enhancement of the lightning channel (at a relatively long time after the establishment of the upward leader) is produced by a recoil leader intercepting the conducting channel to ground, indicating that the M-component is the result of such interception process. They proposed an electrostatic model concerning the concept of bidirectional leader and argued that for M-component, there could be no charge transfer by a downward wave inside the conducting channel to ground. Though the statements of Rakov et al. [1995, 2001] and Mazur and Ruhnke  are both about the mechanism of the M-component, their focuses are actually different. Rakov et al. [1995, 2001] are concerned more about the evolution of M-component through the channel, while Mazur and Ruhnke  paid more attention to what lead to the occurrence of M-component. Here we examine some controversial issues based on our measurements and the observational facts described above. For the issue on whether there is a downward transferring of net charge through the channel, Mazur and Ruhnke  kept an adverse opinion; however, our measurements support the existence of it, consistent with Rakov et al. [1995, 2001]. Our data from triggered lightning showed that at close distances, the leading edge of the E-field pulse by M-component is somehow similar to that by dart leader-return stroke (refer to Qie et al. ), although the risetime of the leader field was considerably shorter than that of M-component. As in Figure 3, the negative changing of the E-field indicates an approaching of negative space charge to the E-field sensor, that is, a lowering of the negative charge down through the channel. Jiang et al.  once simulated the E-field changes associated with the M-component on the basis of the two-wave theory and the measured channel base current and confirmed that such theory is generally appropriate for interpreting the evolution of the M-component through the channel, although some modifications are needed. The assumption (by the two-wave theory) that both the downward wave and the upward reflected wave are unchanged while propagating in the channel may not be very suitable, because their mutually superimposition will certainly lead to an interaction which impacts each other, and hence, the upward wave should not just be the mirroring of the downward one without any change. Nevertheless, the two-wave theory explains the electromagnetic fields at different distances well and provides a valuable understanding of the evolution of charge transferring process in the channel. The modified mechanism of M-component introduced in the following section is mainly based on this theory, together with the existing observation facts by different authors.
4.2 A Modified Mechanism of M-Component That Transfers Negative Charge to Ground
 Based on the measurements from the triggered lightning experiment and the discussion in section 4.1, it can be clarified that M-component contained a downward negative charge transporting wave through the channel, resulting in a negative changing of the electric field at the ground level. For the issue on whether there is a responsive wave propagating upward when the downward wave reaches the ground, an affirmative reply is obtained with responding to the boundary condition, as also mentioned by Winn et al. . Meanwhile, the simulation conducted with the assumption that there is not any responsive upward wave exhibits unacceptable disagreement between the modeled E-field and the measured E-field (as also discussed by Rakov et al. , the overall field waveform could not be well reproduced in case the reflection coefficient is smaller than 0.95).
Shao et al.  suggested that M-component is primarily a downward propagating or “forward” phenomenon. This may be because the triggering mechanism of M-component is the intracloud breakdown development and it is propagating into and down through the conducting channel, leading to dominance of the downward process during the occurrence of M-component. By analyzing the electric field pulse at 45 km which began prior to the current waveform, Rakov et al.  also considered that an in-cloud process may lead to the initiation of the M-component. Actually, when the complex evolution of the intracloud streamer or breakdown which may need additional observations for a more detailed depiction is disregarded, the intracloud process can be roughly viewed as the source of the negative charge supplying into the channel, and hence, sustains the propagating of the downward wave. So here we mainly focus on the evolution and interaction of the M-component waves (propagating downward and upward) through the lightning channel.
 Considering the relationship between the charge transferring process in the channel and the resultant electric field (essentially electrostatic field), the influence of the charge at lower channel to the E-field will be dominant at the very near distance, while that of the charge at higher channel could be negligible. As the horizontal distance increases, the charge at the higher channel contributes more to the E-field. The narrowness (compared with the width of the current waveform) of the E-field pulse at very near distance, as shown in Figure 4, demonstrates that it is in a quite short time than the lower part of the channel being influenced by the negative charge, and after such influencing, although the lightning channel is still carrying a considerable current that can be viewed at the channel base, the lower channel section is most likely neutral, or at least, has a small line charge density. This fact indicates that as the upward wave develops and superimposes to the downward wave, the negative charge previously transferred (from the cloud to the channel) by the downward wave will be basically (or for the most part) depleted. Figure 5 depicts schematically the modified mechanism of M-component that transfers negative charge to ground, and the main points of this mechanism are as follows:
 There is a continuing current-carrying channel connected to the ground. The continuing current is steady and does not contribute to an abrupt increasing or reducing of net charge in the channel.
 The downward wave propagates from the top of the channel toward the ground, transporting negative charge through the channel.
 When the downward wave reaches the ground, the upward wave starts to develop and drains the negative charge transported by the downward wave (or equivalently, realizes the neutralization). As the neutralization occurred, the channel beneath the front of the upward wave is well conductive, with the potential essentially being the same with that of the ground. The potential difference of the upward and the downward waves drives the interaction between them, and the upward wave serves to deplete the negative charge by the downward wave at the interface of them and makes the charge density of the channel beneath the interface layer to be roughly zero; then ideally, the channel base current is equal to the current at the interface of the downward and upward waves.
 According to this modified mechanism, the M-component is modeled based on the following:
 Assume that t0 is the start moment of an M-component current at the channel bottom, which is also the moment at which the downward wave reaches the ground and the upward wave starts. At t* (t*≥ t0), the height of the interface of downward and upward waves is as follows:
where v2 is the propagating velocity of the upward wave. At the same time, the line charge density of the channel beneath H* is
where h is the height of a certain point of the channel.
 The channel base current is equal to the current at the interface of the upward and downward wave (concerning a space charge but not the effect of a displacement current):
where v1 is the propagating velocity of the downward wave, and for this wave, considering that the charge at a height of H* (at the moment t*) was lowered from the height of hm (at the moment t0) during t* − t0, then there are
 Based on the measured channel base current and formula (5), the line charge density ρL(h,t0) at any height of the channel at the moment of t0 can be simulated. Furthermore, the spatial and temporal distribution of ρL is as follows:
 The corresponding vertical electric field (essentially electrostatic field) at ground level above a perfectly conducting earth is as follows:
where H is the vertical length of the lightning channel, D is the horizontal distance from the observational site to the lightning channel, and .
 We assume that the upward and downward waves involve the same propagation speed (v1 = v2) with the order of magnitude of 106–107 m/s, which is reasonable according to the optical measurements and VHF imaging of the M-component [Jordan et al., 1995; Shao et al., 1995]. The length of the lightning channel (H) is hypothesized to be 4 km, which is sufficient to take into account virtually all channel sections contributing to electric field within several hundred meters.
 Figure 6 shows the simulated E-fields at different distances from the lightning channel, based on the modified M mechanism and the measured current (with the peak value of 1.04 kA). The calculated and measured electric field waveforms are in good agreement, indicating the applicability of the model.
 The rational parts of the two-wave theory by Rakov et al. [1995, 2001] have been adopted as the basis of the model for M-component. The main modification lies in the assumption (3) mentioned above, which has roughly defined the interaction between the downward wave and the upward wave and treats their evolution as a whole (to some extent), then hence, contributed to the improvement of the modeling results showing better matching of the calculated and measured E-fields. In the case of the downward wave being analogous to (or acting as the role of) a dart leader while the upward wave being accordingly analogous to a return stroke, the neutralization (as equivalent to draining the negative charge transported by the leader) should not be neglected. Such neutralization interprets well the E-field measurements by Miki et al. : The lower channel section is neutral soon after the downward wave reaches the ground, leading to the narrowness of the E-field pulse at a very near distance. Note that in the two-wave theory, the superposition of the incident wave and the reflected wave with contrary polarities (the reflection charge density coefficient is −1) may be overall viewed as a neutralizing process. However, since there is a time shift between the insert and reflected waves, the reflected wave will dominantly occupy the lightning channel after the main pulse of the M-component, equivalent to a reversal charge surplus (in accordance with the defined reflection charge density coefficient) in the channel, which causes an unexpected overshooting at the railing edge of the simulated E-field (see Figure 12 in Rakov et al.  and Figure 4 in Zhang et al. ). It should be additionally stated that the modified model here has sketchily neglected the wavefront shapes and the junction area between the downward wave and the upward wave should not just be a sharp boundary but actually involves a certain channel length, for not exhibiting a discontinuity in line charge density. The E-field measurements at extremely close distance from the channel have not been conducted in SHATLE experiment, so in Figure 6b, just the simulated E-field (at 1 m) is shown. It is clear that the simulated E-field at 1 m involves a considerably smaller pulse width than that at 60 and 550 m, which is reasonable according to the above analysis. Compared to the records by Pockels sensor as shown in Figure 4, the pulse width in Figure 6b is larger, possibly due to the longer risetime of the current waveform in our case. As in Figure 4, the leading edge of the current waveform is much steeper than the trailing edge. By further comparing Figure 6b with Figure 4, it seems that the predicted field magnitude is a little smaller than the measured one, if the current peaks are taken into account (proportionally). However, as also reported by Miki et al.  that 8 out of 10 M-components involved no significant electric field features by Pockels sensors (with lower detecting limit of 20 kV/m), we consider the field magnitude in Figure 6b to be rational. It is necessary to note that the propagating speed of the waves is an adjustable parameter of the modeling, and for a same case, a larger feeding speed value could result in a smaller simulated electric field magnitude at a certain distance. Figure 7 shows the simulation of electric fields (with the distance of 60 m) for two cases with the peak currents of 6.7 and 1.1 kA, respectively. The M-component shown in Figure 7a has a current magnitude approximately 6 times larger and an E-field magnitude approximately 4 times larger than the M-component in Figure 7b, and the suitable speed value for the simulation of M-component shown in Figure 7a is larger than that in Figure 7b. It is of worth to generalize that for the current pulses of M-component mode transferring negative charge to ground, those with larger magnitude ratios of electric field (at a certain distance away from the channel) to current may involve smaller propagation speeds, possibly determined by the conductivity of the continuing current discharge channel.
 Recently, Winn et al.  analyzed the luminous pulses during a triggered lightning (in New Mexico) with multibranches within the field of vision of the high-speed camera and found that the complex luminous pulses observed in the main channel to the ground were the result of the previously origination of a luminous pulse on dark branch and then it is merging into a continuing luminous branch. This observation is in good agreement with the finding of Shao et al. . They further argued that the dividing of the luminous pulses into two modes of charge transfer by Flache et al.  (as introduced in section 4.1) is unreasonable. Here, combined with the observations by different authors [e.g., Shao et al., 1995; Rakov et al., 2001; Mazur and Ruhnke, 2011; Yoshida et al., 2012], we agree with Winn et al.  on the issue of the generation (or origin process) of M-component. Nonetheless, it should also be emphasized that the modes of charge transfer by M-component and dart leader-return stroke are different, mainly concerning the findings in triggered lightning. As already been clarified by Qie et al. , the pattern of the channel base current versus E-field at close distance for dart leader-return stroke is different from that for M-component. The start of the return stroke current corresponds to the instant of the E-field peak caused by the whole leader-return stroke process, while for M-component, there is a time lag between the E-field peak and the start of channel base current, indicating a sustained transporting of the net charge from the upper to the lower channel after the downward wave reach the ground. Since only the currents were shown in Flache et al. , it is not able to determine the patterns of current versus E-field for fast and slow pulses. However, we tend to think that their fast pulses with very short risetime were correlated with the dart leader-return stroke mode of charge transfer while the slow pulses with the M-component mode, although Winn et al.  pointed out that both pulses may similarly be due to a dart leader merging into a conducting channel. Actually, if those slow pulses associated with continuing luminous branch were due to an intercepting of a leader occurred above the view of the camera, then correspondingly, the channel between the junction point and the lower extremity may be considerably longer than that of the fast pulses with newly illuminated braches within the view of the camera. So the length of the channel section beneath the junction point (flowing with continuing current) may act as an important role in the physical process of M-component. And being long enough of such channel section is a basis of the modified model discussed above.
 The statistical characteristics of M-components following the return strokes of negative rocket-triggered lightning flashes are analyzed in detail. The GM values of the peak current, half peak width, and charge transfer were 276 A, 242 µs, and 101 mC, respectively. Out of 63 M-components, 11 exhibited peak currents exceeding 1 kA with sharper waveforms and higher charge transfers than the usual ones with peak current smaller than 1 kA. The negative-going change of the electric field of M-components started prior to the emergence of the channel base current from the background level, and after the emergence of the current, such electric field was still going larger. There was a time lag between the peak of the current at the channel bottom and the peak of the associated electric field. Several authors have reported the characteristics of M-components [e.g., Thottappillil et al., 1995; Rakov et al., 1995, 2001; Qie et al., 2011]. Our observation results here are statically consistent with the previous results.
 Based on the existing observational facts of M-components and the observed behaviors of current and electric field waveforms, a modified mechanism of M-component that transfers negative charge to ground is proposed. The modified mechanism confirmed that the evolution of M-component through the lightning channel involves a downward wave transferring negative charge from the upper to the lower channel and an upward wave draining the charge transported by the downward wave. And the charge density of the channel beneath the interface layer is roughly zero. The modeling results demonstrated a good agreement between the simulated and the measured E-field waveforms, and very good behaviors of current and electric field waveforms of M-component are reproduced.
 Based on the modified mechanism and the discussion above, the observation facts by different authors can be interpreted in a reasonable way to depict M-component. A downward wave caused and driven by in-cloud discharge processes (or streamer/leader developments) acting upon the upper extremity of the lightning channel flowing with the continuing current could conduct negative charge transportation from the upper channel to the lower channel, and hence, result in the negative changing of the electric field at ground level. The upward wave drains the negative charge transported by the downward one, acting like a return stroke with considerably smaller propagating speed (than return stroke). The considerably long conducting channel with continuing current has influenced the leader-like downward wave and the return stroke-like upward wave, leading to a different pattern of charge transferring mode (from cloud to ground) from dart leader-return stroke.
 The research was supported by National Natural Science Foundation of China (grant 41175002, 40930949) and One Hundred Person Project of the Chinese Academy of Sciences. Appreciation is owed to all the participants taking part in the SHATLE experiment.