According to the substorm paradigm, the tail magnetic flux F is accumulated during the growth phase and suddenly released afterward. However, only 33 out of 142 isolated midtail sudden unloading events identified by Geotail in 1995–1998 demonstrated monotonous F increase (type 1 substorms). In 32 cases, balanced tail flux (BTF) period (F ~constant) with enhanced dissipation and pseudobreakup activity during 0.5–2 h before unloading was registered (type 2, a new substorm class). The rest 77 substorms lie between types 1 and 2. Superposed epoch analysis revealed that during ~40 min before unloading, both groups demonstrate almost identical plasma sheet thinning and Bz decrease at Geotail. By the sudden unloading time, both groups have nearly identical Bz values in the midtail as well as at 6.6 RE. We conclude that pre-onset midtail configuration changes are common for all substorm types irrespective of flux growth.
As established in the early 1970s, isolated substorms are preceded by a growth phase [Russell and McPherron, 1973], which is now a part of the commonly accepted substorm paradigm. It suggests that during that phase the magnetic flux provided by enhanced dayside reconnection is transported to (and accumulated in) the magnetotail. This unbalanced flux transport leads to an increase of the magnetotail flaring and of the lobe magnetic pressure, plasma sheet thinning, and the tail field stretching. The paradigm suggests that these changes eventually result in some instability, initiating a sharp current disruption and rapid magnetic flux unloading (that is, the substorm onset). The most easily observed growth phase effect in the tail is an increase of the total magnetotail pressure, reflecting the tail magnetic flux and pressure buildup, as confirmed in many studies (starting from Caan et al. ).
However, deviations from this simple behavior have also been reported in the literature, including the widely discussed problem of how to objectively define the substorm onset and, in particular, the problem of pseudobreakups [e.g., Kullen and Karlsson, 2004]. The role of the magnetotail field as the main reservoir of energy dissipated during substorms have been questioned by Pulkkinen et al. . Several examples of substorm onsets, not preceded by a pressure or magnetic flux increase in the magnetotail, have been presented [Petrukovich et al., 2000; Yahnin et al., 1994, 2001; DeJong et al., 2008]. This has also been statistically confirmed by Dmitrieva et al.  who noticed that ~1/3 out of 145 isolated substorm events in the midtail studied did not show any total pressure growth before unloading, whereas a classical growth phase was observed only in the other 1/3 of events (the remaining 1/3 showed a mixed behavior).
These results indicate that the magnetic flux and pressure growth in the magnetotail may not be a necessary precondition for the sudden release of tail magnetic energy and call for further studies to answer the questions “What is common in different types of the magnetic flux behavior?” and “What are the critical conditions for the sudden unloading?” It is also important to understand what is going on in the magnetotail during stable tail magnetic field interval before the sudden magnetic energy release.
This paper attempts to address these questions. To that end, we extend the study of Dmitrieva et al.  in a few respects. We identify the substorm onsets based on their sudden unloading signature observed in the midtail (which is the earliest and a robust onset signature, see Miyashita et al. ). Our main data source is Geotail (GT), but instead of estimating the total pressure we use the magnetotail magnetic flux as an adequate quantity in the flux balance considerations; we also consider plasma velocities at GT. In addition, we also use near-midnight geosynchronous observations to probe the dipolarizations and field stretching in the inner tail, as well as ground-based data.
2 Data Analysis
For years 1995–1998 we used simultaneous Geotail magnetic and plasma data in the middle tail (−32RE<X<−15RE, |Y|<15RE) and Wind solar wind data, propagated to Geotail X position, to calculate the magnetotail magnetic flux (F) using the method described by Shukhtina et al. . According to the “calibration” provided by Shukhtina et al. [2009, 2010] the calculated F values exceed the “real” magnetotail flux values (obtained from MHD simulations and global auroral images) by ~0.2 GWb due to finite plasma sheet thickness, so in the following, this offset is subtracted from the calculated F values. The magnetic flux calculation requires knowledge of the solar wind dynamic pressure (Pd=1.94nV2, assuming 4% helium content), of its static pressure, as well as GT coordinates and the equivalent lobe magnetic field BL at Geotail location, calculated from the pressure balance BL=(B2+2μ0nkT)1/2 in the midtail.
To characterize the solar wind-dependent dayside reconnection rate, we use the so-called “merging electric field” Em=VBT sin3Θ/2, where and Θ is the IMF clock angle, calculated from the propagated Wind measurements. At geosynchronous orbit we use GOES 8 and GOES 9 magnetic field data in the ±3 h magnetic local time sector to characterize the magnetic field stretching in the inner magnetosphere. When computing the tail flux we used 6 min averaged data to minimize possible timing mismatch between Geotail and propagated Wind data. For other GT parameters, 2 min averages were used, whereas Bz values at 6.6RE as well as ground AL index data have original 1 min resolution.
To identify sudden unloading events we used the criterion ΔF/Δt<−0.1 GWb/10 min) and defined the unloading time moments, TU, at 2 min resolution. Then we checked if the sudden unloading was accompanied (within 15 min) by an AL intensification (AL<−100 nT). Finally, we selected 142 intense isolated substorms without sharp unloadings during the preceding hour.
Next, we performed the analyses in different subgroups. From the magnetic flux circulation viewpoint, there are two extreme types of possible behavior: (1) all magnetic flux transported from the dayside is accumulated in the tail, and (2) the day/night reconnection rates are equal, with no flux accumulation. For the first group, we selected 33 ideal substorms with monotonous flux increase from the quiet background with ΔF>0.5 (<Fonset>−<FSMC>). The statistical levels <Fonset> and <FSMC> just before onset and during the steady magnetospheric convection (SMC) events were obtained by Shukhtina et al.  as a function of solar wind parameters. Typically, this threshold value was about 0.2 GWb. These type 1 events constitute 23% of the total number of substorms. The tail flux F in this group, on the average, grows from 0.52 GWb to 0.83 GWb, or by 60%, during the growth phase of an average duration of 55 min. For the second extreme group, we required that the F changes between TS and TU were less than 0.1 GWb (average |F(TU)−F(TS)|= 0.03 GWb). Such magnetic flux dynamics contradicts the standard substorm paradigm, and we define it as a new substorm class (type 2). We found 32 substorms of this type or 23% of all events. A list of parameters of type 2 substorms is available in Table S1 in the supporting information.
Figure 1 presents results of the superposed epoch analysis for these two marginal groups with TU being the zero epoch. According to the figure the average duration of the balanced tail flux (BTF) phase from the flux stabilization time TS until TU is 36 min (varying between 15 and 120 min) with the average F level 0.72 GWb. The AL index drop near TU confirms that our unloading-based timing is consistent with the standard AL-based method of substorm onset determination. The figure shows that type 2 events are characterized by slightly more active electrojet conditions than type 1 events. The dynamic pressure Pd is stable and has the same value (2.8 nPa) for both types (not shown). Note that the solar wind driving (Em value) is nearly the same for both groups during the BTF period but considerably differs before. This enhanced driving during the preceding time caused the tail flux accumulation and magnetotail stretching to occur earlier for type 2 events. By the time t=TU−1 h Bz at GOES is <60 nT for type 2 substorms compared to Bz>70 nT for type 1 events. In this sense, the type 2 events also had a growth phase, separated from the expansion (unloading) phase by the BTF period, so we define them later as the “delayed onset” events. The 36 min interval between TS and TU is the most interesting time period for comparison. The solar wind driving conditions are the same for both groups, but the magnetotail consequences are quite different. The growth phase is in progress for type 1 events, whereas the tail flux stays balanced for type 2. In the near tail, Bz decreases at GOES for type 1 substorms, being saturated for type 2 and coming to the same value for both groups by the time TU. In the midtail, Bz decreases for both types in a similar manner, and the beta parameter also behaves similarly, indicating plasma sheet thinning. By the time TU, Bz values for two event types are nearly the same in the midtail as well as at 6.6RE. Based on these characteristics, the magnetotail (including near and middle tail) evolves to similar configurations at zero time, starting from different configurations 35 min before. However, the tail configurations at zero time are not fully identical: the tail magnetic flux is 15% larger in ideal events (0.83 against 0.72 GWb), whereas the Bz/BL value is larger and plasma beta is smaller for delayed onset events.
The BTF phase implies that the nightside reconnection and earthward plasma convection provide the flux return and enhanced dissipation in the tail plasma sheet and in the ionosphere. This is clearly seen in the average flux transfer rate (Ey= −[V×B]y) at GT in Figure 1, which shows enhanced <Ey>=0.29 mV/m during the BTF phase, in striking difference with small <Ey>=0.06 mV/m during the growth phase of ideal events. Also, close to TU, one may notice indications of a weak electrojet activation (small-AL decrease) and small dipolarizations (positive Bz variations) at GOES and GT, both occurring at time TS±5 min (purple triangles). During the unloading, the solar wind driving Em stays at the same level for type 1 events but shows some decrease for type 2 substorms. Though it could indicate more effective external triggering for type 2 events, this conclusion needs a more thorough study.
As an example of type 2 substorm and BTF phase, we provide in Figure 2a summary of an isolated substorm on 16 November 1995. We refer to Yahnin et al.  for details of comprehensive auroral, geomagnetic, and geosynchronous observations. Though there are no OMNI data for this period, Wind and IMP 8 data exist, making it possible to propagate it to Geotail position and calculate the magnetic flux. As there are gaps in IMF data, as a proxy of Em, we used PCS index (from sunlit polar cap, Figure 2 (top)), which indicates that the convection increase started after 2220 UT and continued throughout the whole event. As shown by Yahnin et al. and confirmed by magnetic flux increase in Figure 2, the growth phase was in progress for the next 45 min. Starting from 2307 UT and until the substorm onset at 2336 UT, a new phase was identified. Yahnin et al. called it as “pseudobreakup phase,” based on observations of bright auroral breakups, PiB pulsations, and very soft plasma injections at Los Alamos National Laboratory (LANL) spacecraft in the presence of very weak magnetic perturbations observed underneath of the bright auroral displays (IL index changes between −50 and −150 nT; see Figure 2). Based on the observed approximately constant magnetic flux and BL at GT, confirmed by the Interball observations, we would call this 30 min long episode as the BTF phase in our classification. The expansion onset at 2336 UT is very strong in all signatures, including unloading and westward jet intensification. This event illustrates most of our findings and, in addition, provides a direct link of enhanced dissipation during the BTF phase to the well-known pseudobreakup class phenomena.
Two extreme groups, the ideal events and the delayed onset events taken together, constitute about a half of our data base. The rest of events are expected to lie between the two extremes, and our statistical results in Figure 3 confirm this expectation, indicating that the magnetotail evolution toward the sudden unloading includes a continuum of different pathways, bounded by the two extremes (types 1 and 2 events). Note that Bz values at TU at 6.6RE as well as in the midtail are nearly the same under the same external driving for all three groups.
It is interesting that at the onset time Vx is positive for type 2 events and negative for types 1 and 3, with corresponding different signs (negative and positive) of Bz variation. As the GT position is almost the same for all groups (~ 23–24 RE), it may indicate different reconnection line locations for groups 1 and 3 (earthward of 24 RE) and for group 2 (tailward of 24 RE). This is consistent with weak dipolarization at GOES for type 2 events.
3 Discussion/Concluding Remarks
In this paper we base our substorm identification on the variations of tail magnetic flux, the primary quantity closely related to the magnetotail reconnection, and the loading/unloading of hypothetical energy reservoir for substorms [Baker et al., 1996]. In spite of the doubts raised by Pulkkinen et al.  about importance of this energy source, the tail magnetic flux stored during the substorm growth phase has shown good agreement with the values of reconnected magnetic flux threading the substorm auroral bulge [Shukhtina et al., 2005]. The system behavior, however, is variable.
Interestingly, the type 2 delayed onset substorms are as common in our data set as the ideal type 1 events are. This provides a few lessons to us.
The tail flux growth just before onset is not necessarily required for the sudden unloading.
A common feature for all groups is a continuing field stretching in the middle tail leading to almost equal Bz values at TU in this region. Almost equal Bz values for all groups at the unloading time are also observed at 6.6 RE.
In spite of similar driving level, the sudden unloading for two groups starts at considerably different (although enhanced) magnetic flux levels (0.83 against 0.72 GWb). This implies that there exists no fixed F threshold to initiate the substorm instability, in agreement with previous conclusions of Shukhtina et al. , DeJong et al. , and Milan et al. .
According to observed Vx and Bz variations the average substorm origin location is earthward (tailward) of 24 RE for type 1 (2) correspondingly.
The BTF phase, separating the loading and unloading phases, in some respects is similar to such activity types as pseudobreakups and steady convection (SMC) events [Sergeev et al., 1996; Tanskanen et al., 2005]. Similar to the SMC state the BTF periods are characterized by balanced dayside and nightside reconnection rates but differ from SMC by shorter duration (~0.5 h, compared to >3–4 h in Sergeev et al.  and >1.5 h threshold used, e.g., by Kissinger et al. ) and slightly smaller AL. The BTF phase of delayed substorms can thus be viewed as the short-duration and low-AL extension of the SMC-like activity. One more significant quantitative difference is the average tail magnetic flux value during the BTF phase (0.72 GWb), considerably higher than <FSMC>=0.57 GWb from DeJong et al.  and <FSMC>=0.52 GWb obtained for given solar wind conditions by Shukhtina et al.  and shown in Figure 3. Compared to BZ numbers in the latter study the magnetotail configuration during BTF is more stressed compared to regular steady convection events.
To conclude, we show that besides ideal growth phase substorms there exists a subclass of numerous substorms with delayed onsets. This subclass displays balanced tail flux (BTF) interval (>0.5 h) before the sudden unloading which presumably results from reconnection tailward −24RE as distinct from ideal substorms. A numerous event group 3 shows the intermediate behavior between the ideal and delayed onset substorms extremes. Enhanced plasma sheet convection and bursty dissipation similar to pseudobreakups are common for BTF. These BTFs are somewhat similar to short-duration SMC events but differ from the latter ones by configurational changes (identical to usual substorms), stretched magnetotail configuration, and larger magnetotail magnetic flux level.
Excellent Geotail data have been provided via ISAS DARTS website, and solar wind/activity data were available from NASA/GSFC SSCWeb and OMNI (SPDF OMNIWeb) databases; we thank the experiment teams and data providers. We also thank A.G. Yahnin and N.A. Tsyganenko for discussions and C. Jacquey for useful suggestions. The work was partly supported by FP7 grant 263325(ECLAT), SPbU grant 22.214.171.1241, and RFBR grant 13-05-00132.
The Editor thanks Christian Jacquey and an anonymous reviewer for their assistance in evaluating this paper.