Statistical study on interplanetary drivers behind intense geomagnetic storms and substorms

Geomagnetic storms and substorms play a central role in both the daily life of mankind and in academic space physics. The profiles of storms, especially their initial phase morphology and the intensity of their substorms under different interplanetary conditions, have usually been ignored in previous studies. In this study, 97 intense geomagnetic storms (Dstmin ≤ –100 nT) between 1998 and 2018 were studied statistically using the double superposed epoch analysis (DSEA) and normalized superposed epoch analysis (NSEA) methods. These storms are categorized into two types according to different interplanetary magnetic field (IMF) Bz orientations: geomagnetic storms whose IMF is northward, both upstream and downstream relative to the interplanetary shock, and geomagnetic storms whose upstream and downstream IMF is consistently southward. We further divide these two types into two subsets, by different geomagnetic storm profiles: Type I/Type II — one/two‐step geomagnetic storms with northward IMF both upstream and downstream of the interplanetary shock; Type III/TypeIV — one/two‐step geomagnetic storms with southward IMF both upstream and downstream of the interplanetary shock. The results show that: (1) geomagnetic storms with northward IMF both upstream and downstream of the interplanetary shock have a clear initial phase; geomagnetic storms with southward IMF in both upstream and downstream of the interplanetary shock do not; (2) the IMF is an important controlling factor in affecting the intensity characteristics of substorms. When Bz is positive before and after the interplanetary shock arrival, the Auroral Electrojet (AE) index changes gently during the initial phase of geomagnetic storms, the median value of AE index is maintained at 500–1000 nT; (3) whenBz is negative before and after the interplanetary shock arrival, the AE index rises rapidly and reaches its maxmum value about one hour after storm sudden commencements (SSC), although the time is scaled between reference points and the maximum value of AE is usually greater than 1,000 nT, representing intense substorms; (4) for most cases, the Dst0 usually reaches its minimum at least one hour after Bz. These results are useful in improving contemporary space weather models, especially for those that address geomagnetic storms and substorms.


Introduction
Geomagnetic storms and substorms are important space weather phenomena and an intensely active space weather research topic. Many investigators have studied geomagnetic storms and substorms (e.g., Akasofu and Chapman, 1963;Lakhina et al., 2006;Russell et al., 1974;Partamies et al., 2011). It is well known that substorms can occur during super storms and intense storms, or even independently of storms (Hajra et al., 2016;Partamies et al., 2013;Tsurutani et al., 2015). Hajra et al. (2016) studied supersubstorm (SSS) events from 1981 to 2012 by using superposed epoch analysis (SEA) and found that 86.5% of the SSS events occurred in the main phase of geomagnetic storms and about 9.5% occurred in the recovery phase. Recently, study of space weather conditions for SSS events has attracted increased attention; it has been noted that SSS events appear to be associated with enhanced val-ues of solar wind speed and with southward direction of the interplanetary magnetic field (IMF) (Tsurutani et al., 2015;Zhou XY and Tsurutani, 2001). Yue C et al. (2010) and Ma XH et al. (2019) reported that interplanetary (IP) shocks with southward IMF preconditions can significantly enhance Auroral Electrojet (AE) indices, while no obvious AE index enhancements are observed when IP shock events occur under northward IMF preconditions.
For geomagnetic storms, many studies have focused on the relationship between the intensity of geomagnetic storms and interplanetary sources and/or on causes of the one-step/two-step morphology of the main phases of geomagnetic storms (Akasofu and Chapman, 1963;Russell et al., 1974;Srivastava and Venkatakrishnan, 2004;Le GM et al., 2010;Zhao H et al., 2011). It is now well established that intense geomagnetic storms (Dst min ≤ 100 nT) are caused mainly by coronal mass ejections (Brueckner et al., 1998;Gonzalez et al., 1999;Cane et al., 2000;Gopalswamy et al., 2000;Zhang J et al., 2003), while moderate and weak storms can be caused by high-speed streams in corotating interaction regions (Gosling et al., 1991;Tsurutani and Gonzalez, 1997). Some re-searchers have classified geomagnetic storms as either one-or two-step by their profiles and have studied the interplanetary magnetic field, pointing out that the steps of the time varying Dst during storms result primarily from different structures of the interplanetary solar wind (Kamide et al., 1998;Dessler and Parker, 1959;Tsurutani and Gonzalez, 1997;Richardson and Zhang J, 2008).
A typical geomagnetic storm has three phases: an initial phase, a main phase, and a recovery phase (Gonzalez et al., 1994). The initial phase is characterized by increases of 20 to 50 nT in Dst (or in its one-minute component SYM-H) in tens of minutes. The initial phase often begins with an abrupt increment of Dst referred to as a storm sudden commencement (SSC). As mentioned previously, most studies of geomagnetic storms have focused on the storms' intensities or main phases. The initial phases of geomagnetic storms under different interplanetary conditions, and corresponding substorm characteristics, are rarely studied.
SEA is a powerful method for studying the time variation of geomagnetic storms. However, Yermolaev et al. (2010a) pointed out that by using only SEA the interplanetary causes of magnetic storm onset cannot be investigated, since the duration of the storm main phase can vary between 2 and 15 hours, the average duration being approximately 7 h (Vichare et al., 2005;Gonzalez and Echer, 2005;Yermolaev et al., 2007Yermolaev et al., , 2010b. In order to remove SEA's drawbacks these investigators developed a double superposed epoch analysis (DSEA) method. Then, by applying the DSEA method, they presented a comprehensive study of plasma and solar wind properties for different categories.
Similar to SEA and DSEA (Yermolaev et al., 2010a, b), normalized superposed epoch analysis (NSEA) is a statistical analysis method for time series with three or more reference times. In this paper, by using the DSEA and NSEA methods, we systematically analyze the properties of the interplanetary magnetic field, solar wind parameters, and the dawn-dusk electric field that are associated with geomagnetic storms and substorms. We focus primarily on the morphology of the storms' initial phase and on the intensity of substorms under different interplanetary conditions, details that have tended to be ignored in previous studies.
This study is organized as follows: first, we analyze four typical geomagnetic storms events: one-step geomagnetic storms on June 26, 1998 / August 06, 1998 with northward/southward IMF in both upstream and downstream regions of the IP shock, and twostep geomagnetic storms on May 4, 1998/May 8, 2005 with northward/southward IMF in both upstream and downstream regions of the IP shock. Then we use DSEA and NSEA to analyze the properties of IP parameters for four typical geomagnetic storms: Type I/Type II (one/two-step geomagnetic storms with northward IMF both upstream and downstream of the IP shock), Type III/Type IV (one/two-step geomagnetic storms with southward IMF both upstream and downstream of the IP shock).

Data Set and Methodology
Geomagnetic storms are caused primarily by sufficiently intense, long-duration, southward IMF, associated with intense and longlasting interplanetary electric fields (Gonzalez et al., 1994;O'Brien and McPherron, 2000;Farrugia et al., 2006). Xie H et al. (2006) pointed out that there is a good correlation among Dst min , B z , and VB z . The correlation coefficient between Dst min and B z dips in longlived geomagnetic storms is 0.77 and 0.79 particularly in the main phases. Wu CC and Lepping (2002) noticed that the occurrence timing of storm intensity is highly correlated with the occurrence timing of minimum B z (maximum VB z ) for a magnetic cloud with field rotating southward to northward.
In this paper, we focus mainly on analyzing the interplanetary magnetic field, solar wind parameters, the dawn-dusk electric field based on 1-min OMNI data (ftp://spdf.gsfc.nasa.gov/pub/data/ omni/high_res_omni/sc_specific/), and geomagnetic indices (http://wdc.kugi.kyoto-u.ac.jp/wdc/Sec3.html). By applying the DSEA and NSEA methods, we studied 97 intense geomagnetic storms events (Dst min ≤ -100 nT) from 1998-2018. We classified these events into two types according to different IMF B z orientations: geomagnetic storms with northward IMF both upstream and downstream of the IP shock, and geomagnetic storms with southward IMF both upstream and downstream of the IP shock. We divide these two types further into two subsets according to how the Dst reaches the minimum the main phase. Type I -one-step geomagnetic storms with positive IMF B z in both upstream and downstream of the IP shock. Type II -two-step geomagnetic storms with positive IMF B z in both upstream and downstream of the IP shock. Type III -one-step geomagnetic storms with negative IMF B z in both upstream and downstream of the IP shock. Type IV -two-step geomagnetic storms with negative IMF B z in both upstream and downstream of the IP shock. In our database, there are 32 Type I storms, 15 Type II storms, 28 Type III storms, 13 Type IV storms and only 9 cases whose type could not be identified. One-step geomagnetic storms represent a typical magnetic storm, consisting of a main phase during which Dst decreases monotonously and a subsequent recovery phase. Two-step geomagnetic storms are characterized by a two-step decrease in the value of Dst. To identify one/two-step storms, we used the criteria innovated by Kamide et al. (1998).
To apply the NSEA method, one of the key issues is choosing reference times (Yermolaev et al., 2010a, b). For Type I and Type III we have chosen as our two reference points the storm onset (time "0") and the time of the Dst index minimum (time "6"). For Type II and Type IV, besides time "0" (the time at which the Dst index reaches its minimum during the first step) and time "6" (the time at which the Dst index reaches its minimum during the second step), we choose the storm onset (time "-6") as another reference point. For all events, the time series between time "0" and time "6" or between time "-6" and time "0" is divided equally into 5 subintervals and the data in each subinterval are averaged. The time scale before the first reference point and after the last reference point remains unchanged. These artificially similar events are then grouped according to epoch time. Finally, we treat the 25 th percentile, the 50 th percentile (i.e., the median) and the 75 th percentile as average statistical properties of the different parameters.

Case Study
For the case study, we have selected four typical geomagnetic storms (one-step geomagnetic storms on June 26, 1998 / August 06, 1998 with northward/southward IMFs both upstream and downstream of the IP shock, and two-step geomagnetic storms on May 4, 1998 / May 8, 2005 with northward/southward IMFs both upstream and downstream of the IP shock) to illustrate the events selected in the four different geomagnetic storms types, which are shown in Figure 1 and Figure 2, respectively. Figure 1 shows the solar wind conditions based on OMNI data, the SYM-H, the Dst, and the AE variation for Type I magnetic storm (Dst min = -101 nT) on June 26, 1998 and Type III geomagnetic storm (Dst min = -138 nT) on August 06, 1998. From top to bottom the panels plotted are the plasma temperature, density, velocity, dynamic pressure, the magnetic field intensity, the dawn-dusk electric field, the IMF B z , the SYM-H, the original geomagnetic index Dst, the pressure-corrected Dst 0 , and the AE index. The light green area represents the initial phase and the purple area repres-ents the main phase of the storm. The red vertical dashed line denotes the time of shock arrival. The time when Dst reaches its minimum is denoted by the sky-blue vertical dashed line. According to Burton's empirical equation (1) (Burton et al., 1975), we calculated the pressure-corrected Dst 0 shown in Panel 9 for each event in Figure 1.
Where E = VB z × 10 -3 (mV/m) and p = nv 2 × 10 -3 (eV/cm -3 ), F(E) is the injection function, B z is the southward IMF, V is the solar wind velocity, p is the dynamic pressure, and the coefficients a and b are constants. The coefficient a, which measures the ring current decay, has the value of 3.6×10 -5 s; the coefficient b, which meas-   Figure 1. Two cases of one-step magnetic storm with northward/southward IMF both upstream and downstream of the IP shock. From top to bottom the panels plotted are the plasma temperature, density, velocity, dynamic pressure, the magnetic field intensity, the dawn-dusk electric field, the IMF B z , the SYM-H, the Dst index, the pressure-corrected Dst 0 , and the AE index. The red vertical dashed line denotes the shock arrival time. The time when Dst reaches its minimum is denoted by the sky-blue vertical dashed line. The light green area represents the initial phase. The purple area represents the main phase. (a) One-step geomagnetic storm on June 26, 1998 with northward IMF B z both upstream and downstream of the IP shock. (b) One-step geomagnetic storm on August 06, 1998 with southward IMF B z both upstream and downstream of the IP shock. ures the correlation between the corrected Dst and dynamic pressure, has the value of 0.20 nT/(eV·cm -3 ) -1/2 .
In Figure 1a, the IMF B z remained positive both before and after the time of shock arrival (at 16:48 UT), as it increased from 8 to 18 nT. Meanwhile, the E y decreased from -3 to -7 mV/m and the SYM-H increased from 20 to 43 nT. In the following interval (between 16:48 and 23:00 UT), B z remained positive and E y remained negative, with no ring current injection, and the higher value of SYM-H persisted; this completed the initial phase. Meanwhile, the AE index remained at 0-100 nT. After 23:00 UT, B z rotated from positive to negative, E y became positive and produced current injection, and thus induced the main phase of the magnetic storm, during which AE increased from 72 to 1232 nT, reaching its maximum, the level of intense substorm, 10 hours after SSC.
The penultimate panel of Figure 1a shows the original Dst and pressure-corrected Dst 0 . The dynamic pressure began to increase from 16:48 UT, with the contribution of magnetopause current leading to Dst index increases, while the Dst 0 basically did not change. After 23:00 UT, the variations of Dst and Dst 0 versus time are similar, and we focus mainly on the variation of pressure-corrected Dst 0 . The peak of the storm is driven by long-playing southward directed magnetic fields (B z < 0), which produced prolonged positive E y , ring current injection. As a result, the main phase began and reached the minimum at 04:00 UT.
For comparison, a one-step geomagnetic storm on August 06, 1998 with southward IMF in both upstream and downstream regions of the IP shock is shown in Figure 1b. In this event, B z decreased from -10 to -18 nT after the shock arrival (at 07:32 UT) and the E y increased from 3 to 7 mV/m. The SYM-H increased from -20 to 0 nT then gradually decreased. The AE index reached the moderate substorm level at the shock arrival, and reached the intense substorm level (AE = 1128 nT) during the magnetic storm. The maximum value of the AE index appeared 1.5 hours after SSC. Between the time of shock arrival and the time of Dst reaching its (a) 1998/05/04 two-step storm, B z from "+" to "+"   minimum (at 11:00 UT), the B z remained negative; E y remained positive, producing current injection that drove Dst to a main phase minimum of -138 nT directly, without an initial phase.
It can be seen from Figure 1 that, under southward IMF condition, the August 06, 1998 event affected the geomagnetic indices more than did the June 26, 1998 event. This observation is consistent with the results of Yue C and Zong QG (2011). The Type I storm in this case study had a clear initial phase while the Type III storm did not. The Burton equation can explain this difference. From the time of shock arrival to the time of Dst min in the Type I case, E y changed from negative to positive, resulting in a clear initial phase and then a main phase. In the Type III case, however, a continuous positive E y produced ring current injection that drove Dst to a main phase minimum of -138 nT. We can also see from Figure 1 that when B z is negative before and after the arrivald of the IP shock, the AE index will rise rapidly. An intense substorm will be triggered in 1-2 hour.
It is worth noticing that the time of minimum Dst 0 is not identical to the time of minimum B z , which again the Burton equation can readily explain; this will be analyzed in detail in the discussion section. As we can see from Figure 1 and the first step of the storm in Figure 2, similar results can be found for the IP shock associated with a given IMF B z variation. In theType II case, the IMF B z remained positive both before and after the time of the first shock arrival, maintaining a higher value of SYM-H and defining the initial phase. At 04:00 UT, B z rotated from positive to negative; E y became positive and produced current injection, and thus induced the main phase of the magnetic storm. In the Type IV case, the IMF B z remained negative both before and after the time of shock arrival, and then B z remained negative until the time at which Dst reaches its minimum (at 03:00 UT). As a result, E y remained positive and produced current injection that drove Dst directly to its main phase minimum of -100 nT, without an initial phase.
The variation characteristics of the AE index were similar to those in Figure 1. During the initial phase of the geomagnetic storm, AE ≤ 500 nT. During the main phase of the magnetic storm, the AE index rose to 500 nT, and the maximum value exceeded even 1,000 nT, triggering the intense substorm.

Statistical Study
In Section 3, four case studies indicate that the properties of interplanetary magnetic field, solar wind parameters, ring current injection, and the Dst and AE indices are obviously different among the different types. In this section, we use DSEA and NSEA methods to examine the characteristics of interplanetary parameters and geomagnetic indices for different types of geomagnetic storm events. Figure 3 illustrates the results of a DSEA for Type I and Type III geomagnetic storms. Figure 4 illustrates the results of an NSEA for Type II and Type IV geomagnetic storms. Black, red, and blue lines mark the 25 th percentile, 50 th percentile and 75 th percentile values, respectively.
In Figure 3a, the median value of IMF remained northward both before and after the time of shock arrival (time "0"), as it increased from 1 to 8 nT. Next, at approximately "3" h, it dropped down to zero and reached its minimum (around -20 nT) near time "6". Meanwhile, the median value of the SYM-H index increased from 0 to 20 nT, and then maintained a higher value in the time range "0-3", leading to the initial phase. In the initial phase, the AE index increased slowly and was less than 500 nT. This is consistent with the results of the case study. SYM-H began to drop down near time "3" and reached its minimum -147 nT near time "6", which caused the main phase. During this period, the AE index rose and remained at 500-1000 nT. The median value of the solar wind speed changed from 400 to 500 km/s, then showed no change in the time interval "0-6". The median values of the plasma temperature, density, dynamic pressure and magnetic field intensity all increased abruptly at the time of shock arrival. The dynamic pressure and magnetic field intensity remained approximately constant in the time interval "3-6". The plasma temperature and density both increased in the interval "0-3", slowly decreasing in the interval "3-6".
As can be seen from Figure 3b, the median value of IMF remained southward both upstream and downstream of the IP shock, as it decreased from -2 to -10 nT and reached -10 nT near time "1". Meanwhile, the AE index rapidly increased from 400 to 1,027 nT. The median value of the SYM-H index increased from -10 to 10 nT at time "0", then decreased from 10 to -133 nT, reaching its minimum at time "6", which caused the main phase. The median value of the solar wind speed changed from 380 to 500 km/s, and then showed small changes in the time interval "0-6". The median values of dynamic pressure and the magnetic field intensity showed similar behaviors as those in Figure 3a. In the time interval "0-6", the median value of density remained high (~20 n/cc) while the median value of the plasma temperature decreased.
In addition to the above differences, we can also obviously find that when B z is positive before and after the IP shock arrival, the AE index changes gently during the initial phase of geomagnetic storms; and, usually, the median value of the AE index is less than 500 nT. When B z is negative before and after the IP shock arrival, the AE index rises rapidly and usually reaches its maxmum an hour after SSC, triggering intense substorms, although the time is scaled between reference points. The result is consistent with the case study in Figure 1.
In Figure 4a, the median value of IMF remained northward both before and after the time of the first shock arrival (time "-6"), as it increased from 1 to 8nT, Meanwhile, the median value of the SYM-H index increased from -20 to 0 nT, P d increased from 3 to 8 nPa, and the solar wind speed changed from 450 to 500 km/s. E y decreased from 2 to -3 mV/m. In the interval time "-6 to -3", the solar wind speed changed slightly, the median value of B z remained positive and E y remained negative, with no ring current injection, which led to a high and relatively unchanged value of the SYM-H index, and the storm entered the initial phase. Around time "-3", B z rotated from positive to negative and reached its minimum close to time "-2". At the same time the E y increased from negative to positive, providing the ring current injection that produced the first main phase and reached its first minimum at time "0". In Figure 4b, the median value of IMF remained southward both before and after time "-6". In the range "-6 to 0", the changes of AE in Type II/Type IV were similar to those in Type I/Type III. Meanwhile, B z remained negative, and the continuous positive E y produced ring current injection that caused the decrease of the median value of the SYM-H index from -30 to -73 nT, with its minimum reached at time "0", signaling the first main phase. In the second step of the storm, all parameters of Type II and Type IV showed similar behaviors as in the first step.
From Figure 3 and Figure 4, we notice that the Type I and Type II storms have a clear initial phase (light green area) while the Type III and Type IV storm do not. This can be explained by the Burton equation. Between the time of shock arrival and the time (or the first time) of Dst min , the continuous positive E y for Type III and Type IV storms produced ring current injection that caused a direct main phase. For Type I and Type II storms, E y changed from negative to positive, which resulted in a clear initial phase when the E y was negative, and then the main phase began after the E y became positive.

Discussion
Ninety-seven intense geomagnetic storms during 1998-2018 were studied statistically using the DSEA and NSEA methods applied to data on the interplanetary magnetic field, solar wind parameters, and geomagnetic indices obtained from OMNI.
Our first prominent research finding is that geomagnetic storms with northward IMF both upstream and downstream of the shock arrival (Type I and Type II storms) have a clear initial phase, while storms with southward IMF both upstream and downstream of the shock arrival (Type III and Type IV storms) do not; this finding is illustrated in Figure 5 (which shows only the IP shocks and driven ICMEs for Type I and Type III). The Burton equation (Burton et al., 1975) can explain this difference. In Figure 6 and Figure 7, between the time of shock arrival and the time (or the first time) of Dst min , the continuous positive E y for Type III and Type IV storms produced ring current injection that caused a direct main phase. In Figure 6a, Figure 7a, and Figure 7c, it can be seen that for a long time after the time of shock arrival, the injection term F(E) is zero, with no ring current injection to cause an initial phase. In Figure 6b, Figure 7b and Figure 7d, it can be seen that the abso-lute value of the injection term is positive; thus, ring current injection dominated this interval. As a result, the main phase began directly.
The second prominent finding is that when B z is negative before and after the IP shock arrival, the AE index rises rapidly, and its maximum value usually exceeds 1,000 nT, triggering a strong substorm, as shown in Figure 5. It is worth pointing out that intense geomagnetic storms and substorms usually have complex and intervening interplanetary drivers (Lee et al., 2007(Lee et al., , 2010Lyons et al., 2008). On the one hand, Lee et al. (2007) has shown many cases in which solar wind changes do not trigger substorms, and the possible reasons includes the lack of insufficient substorm growth phase development prior to the potential triggering change, the inherently nontriggering solar wind change, or the nullifying effect may occur when more than one solar wind quantity change simultaneously. To trigger substorms, the solar wind variations need to reduce the convection strength within the inner plasma sheet. On the other hand, storm-time substorms can be triggered  when the magnetosphere is impacted by a substantial increase in solar wind dynamic pressure (Lyons et al., 2008), and substorms have been seen to occur under northward IMF conditions during the recovery phase of three strong storms (Lee et al., 2010). The results of our study are related to the preconditioning of the magnetosphere by SW and IMF, and since IMF changes are accompanied by simultaneous changes in other SW variables such as density and velocity, it is usually not easy to tell which SW/IMF parameter dominates in determining the magnetospheric phenomenon under consideration, especially when we focus on each case instead of on statistically superposed epoch analysis results.
The third prominent feature is that the time of minimum Dst 0 is not identical to the time of minimum B z . In Figure 1 and Figure 2, the time of minimum B z is 1-5 hours earlier than the time of minimum Dst 0 . For statistical study, the Dst 0 usually reaches its minimum at least an hour after B z , although the time is scaled between reference points. Similarly, Gonzalez and Echer (2005) analysed 64 intense geomagnetic storms and found that the average delay between peak B z and peak Dst values is ~2 h. Zhang JC et al. (2006) compared solar wind features of geomagnetic storm events at solar minimum and maximum by performing SEA on 549 storms and found that the peak time difference is 0.3-1.0 hour between the average IMF B z and Dst min . The time difference also can be explained by the Burton equation (Burton et al., 1975). As can be seen in equation (2), when the absolute value of the injection term F(E), which we averaged to hourly resolution for comparison with our other hourly parameters, is larger than the absolute value of the decay term aDst 0 , the dDst 0 /dt is minus and the Dst 0 index will continue to decrease. Otherwise, if the absolute value of the injection term F(E) is smaller than the absolute value of the decay term aDst 0 , the dDst 0 /dt is positive and Dst 0 will began to increase. The time at which dDst 0 /dt changes its sign is not necessarily the same as the time of minimum B z . This is shown quantitatively in Figure 6 and Figure 7, which give each term in the Burton equation for two cases of the one-step storm (in  Figure 6), and Type I, Type II, Type III, and Type IV storms (in Figure 7). Figure 8 shows the pressure-corrected Dst 0 index and 1-hour step predicted Dst * for Type I, Type II, Type III, and Type IV storms. The black line is Dst 0 , the red dashed line shows the 1-hour step predicted Dst * according to the Burton equation, i.e., we use the pressure-corrected Dst 0 index and solar wind and IMF parameters of the current hour to predict the Dst 0 index of the next hour. It can be seen that the 1-hour step predicted Dst * is consistent with the pressure-corrected Dst 0 for the sixstorm types, demonstrating that the Burton equation can reveal the characteristics of Dst 0 variation and therefore the physical mechanisms of these four types of magnetic storm.

Conclusion
In this study, we systematically analyzed the properties of the interplanetary magnetic field, the solar wind parameters, the dawndusk electric field, and the Dst and AE indices by using DSEA and NSEA methods for four different types of geomagnetic storm. We focused primarily on the morphology of the initial phase of the storms and the intensity of substorms under different interplanetary conditions, details that have generally been ignored in previous studies. Our conclusions are summarized as follows: (1) Geomagnetic storms with northward IMF both upstream and downstream of the IP shock (Type I and Type II storms) have a clear initial phase; geomagnetic storms with southward IMF both upstream and downstream of the IP shock (Type III and Type IV storms) do not. The continuous positive E y on both sides of the IP shock for Type III and Type IV storms produces ring current injection that directly causes a main phase. The main phase begins immediately after the compression. In Type I and II storms, however, E y changes from negative to positive after the planetary shock, which results in a clear initial phase while the E y is negative; the storm main phase is initiated only after the E y sign changes to be positive.
(2) The IMF is an important controlling factor in affecting the intensity characteristics of substorms. When B z is positive before and after the IP shock arrival, the AE index changes gently during the initial phase of geomagnetic storms; usually, its median value is maintained at 500-1000 nT.
(3) When B z is negative before and after the IP shock arrival, the AE index rises rapidly and reaches its maxmum value approximately 1 h after SSC, although the time is scaled between reference points. The maximum value of AE is usually greater than 1,000 nT, triggering intense substorms.
(4) The time of minimum Dst 0 is not identical to the time of minimum B z . For most cases, Dst 0 usually reaches its minimum at least one hour after B z does. The variation trend of B z is highly consist- (b) 1998/08/06 one-step storm, B z from "−" to "−" (a) 1998/06/26 one-step storm, B z from "+" to "+" ent with the morphology of geomagnetic storms, and the time at which B z reaches its minimum is at least 1 hour ahead of the time at which the magnetic storm (i.e., Dst index) reaches its minimum. Therefore, B z can be used as an important input parameter for geomagnetic storm modeling to predict storm morphology change and intensity and can be helpful in improving the forecast accuracy of geomagnetic storm models.