Magnetic Signatures of Ionospheric Disturbance Dynamo for CME and HSSWs Generated Storms

Ionospheric disturbance dynamo is one of the main processes that causes perturbations in the upper atmosphere during a magnetic storm. We present a new method, based on the least square fitting, for estimation of the magnetic signatures associated with ionospheric disturbance currents. Using a wavelet semblance analysis, the durations of disturbance dynamo electric fields have been investigated at three longitudinal sectors. For that we have analyzed the disturbance dynamo (Ddyn) for 19 magnetic storms. It has been found that during CME generated storms magnetic signature of Ddyn may be observed—depending on strength of the storm as well as on the duration of interplanetary magnetic field (IMF) Bz southward—in one, two or all three longitudes. The Oscillatory behavior of IMF Bz during the high‐speed solar wind streams (HSSWs) generates Ddyn globally and the corresponding effects are observed at all low latitude magnetic observatories. In this regard, the Joule heating estimation shows that CME and HSSWs generated storms have very different patterns. The Ddyn duration is found to be maximum for the storms occurring during equinox season. Moreover, the HSSWs events are more likely to cause—because of the oscillatory IMF Bz—long lasting Ddyn as compared to CME generated counterpart. This study presents a detailed analysis of disturbance dynamo as affected by longitudinal and seasonal variations. In this regard the difference in magnetic signatures, of CME and HSSWs originated storms, have been highlighted.

With increasingly powerful magnetometers the measurements of the magnetic field, on Earth as well as on board satellites, have become more accurate and the corresponding magnetic data are interpreted in terms of equivalent electric currents. Incoherent scatter radars allow the measurements of ionospheric electric currents (Mazaudier, 1982;Mazaudier & Bernard, 1985;Mazaudier & Blanc, 1982), for example, the ones situated at Chatanika in the auroral zone (Brekke et al., 1974), in middle latitudes at Millstone-Hill/USA and Saint Santin/France (Carpenter & Kirchhoff, 1975;Mazaudier, 1982;Mazaudier & Bernard, 1985) and in low latitudes at Arecibo/USA (Harper, 1977). As such measurements are rare and very expensive, the magnetometer has remained a useful instrument for describing the real electric currents.
Using the theory of the ionospheric regular dynamo (Chapman & Bartels, 1940) it is possible to understand the regular variation of the Earth's magnetic field as generated by the circulation of neutral atmosphere in the dynamo region (90-150 km). The positive ions are preferably driven by the collisions with neutrals which thus have a speed quite different from electrons, which are under the sole influence of a Lorentz force. Consequently, there is a creation of electric currents in the ionosphere (Balfour Stewart's hypothesis; Stewart, 1880).
During a magnetic quiet period, the ionospheric regular dynamo is the origin of the Solar quiet (Sq) and Equatorial Electrojet (EEJ) current systems, respectively at middle and equatorial latitudes. Whereas for the periods of a robust magnetic activity, strong ionospheric electric currents develop in the auroral zone. These currents-named as auroral electrojets-dissipate energy by the Joule heating effect, which in turn modifies the temperature, pressure and movement of the thermosphere. The disturbed thermospheric winds spread toward mid and low latitudes and, by the dynamo effect, create disturbed ionospheric electric fields, that is, DDEF. Blanc and Richmond (1980) presented the first numerical simulation of the Ionospheric disturbance dynamo. In the present study we have focused on the magnetic signatures of this physical process. Mazaudier (1985) observed the disturbance winds-due to Joule heating-by using the data of incoherent scatter radar, their findings are consistent with predictions of the Blanc and Richmond model (1980). Another solar wind and/or magnetosphere activity associated with auroral electrojets, and results in simultaneous penetration of ionospheric electric fields from high to low-latitudes at all longitudes, is the so-called prompt penetration of the electric field (PPEF) (Nishida, 1968). Vasyliunas (1970) established the first model of the penetration of magnetospheric convection by reproducing the two current cells of the DP2 equivalent current system. There are also electric currents in the magnetosphere (Cole, 1966;Fukushima & Kamide, 1973) which influence the magnetic observations around the globe and one needs to remove the effects of these currents while studying ionospheric dynamics. Thus, the magnetic signature of the disturbed ionospheric dynamo must be extracted from a complex magnetic signal, which is an integrated effect of different ionospheric and magnetospheric electric current systems circulating in the terrestrial environment. Fejer et al. (1983) were the first to point out the effect of PPEF and DDEF on incoherent scatter radar data of Jicamarca. They observed the electric field due to the penetration of the magnetospheric convection (PPEF) on the day of a magnetic storm, and the electric field due to the disturbed ionospheric dynamo (DDEF) during the recovery phase of a storm. Likewise, Le Huy and Amory- Mazaudier (2005) in their study selected the cases having minimum auroral activity during the recovery phase of storm and thus highlighted the D dyn magnetic signature of the disturbed ionospheric dynamo. Zaka et al. (2010) presented the latitudinal variations of disturbance dynamo generated during the two events of 1993 and compared the results with DP2 disturbance. Fathy et al. (2014) analyzed the longitudinal variation of disturbance dynamo during the coronal hole event of April 2010. Fejer et al. (2017) summarized the recent progresses made to analyze the disturbance dynamo and its effect at middle and low latitudes. Nava et al. (2016), Zaourar et al. (2017), Bulusu et al. (2018) and Younas et al. (2020) studied the temporal variation of disturbance dynamo generated during different storms by applying the band pass filter techniques. However, the D iono can have Sq like variations, which may not be considered as an effect of DDEF. Hence, applying simple band pass filters to D iono may overestimate the duration of DDEF. Here, we have applied a new method, based on least square fitting, for the calculations of ionospheric disturbance current. Using a wavelet-based semblance analysis, we have estimated D dyn by separating only those periods which show truly anti-Sq variations. For that we have analyzed 19 different magnetic storms of various origins.
The rest of this article is organized as follows; Section 2 describes the methodology used in this study, while Section 3 presents five events of different categories. In Section 4, we have discussed D dyn variations with Joule heating. The summary and conclusion of the study is presented in Section 5.

Interplanetary Data
Two of the key parameters, namely interplanetary magnetic field (IMF) and solar wind speed, characterizing the space weather activities are provided by Advanced Composition Explorer (http://www.srl.caltech. edu/ACE/) via OMNIWEB data center (https://omniweb.gsfc.nasa.gov/).

Magnetic Indices
The SYM-H, ASYM-H, and AE data is obtained from World Data Center for Geomagnetism, Kyoto Japan. Here SYM-H and ASYM-H indices are, respectively the estimation of symmetric and asymmetric part of storm-time ring currents, while AE is used to assess auroral currents (http://wdc.kugi.kyoto-u.ac.jp/).

Magnetic Observatories and Magnetic Data Analysis
For all the considered 19 magnetic storms, we have analyzed the magnetic data of observatories located at low latitude in Asian sector (Guam, GUA), African region (M'bour, MBO) and American sector (Kourou, KOU). For some of the cases we have also included the observations of Adis Ababa (AAE) and Bac Lieu (BCL). The geographical locations and coordinates of the considered magnetic observatories are shown, respectively in Figure 1 and Here D mag is the disturbance due to magnetospheric currents and o H H H    denoting the change in H, after subtracting the corresponding five hours average of local midnight value H o . The disturbance in the H at low latitudes is mainly influenced by the zonally symmetric ring currents-whose strength is estimated by the high resolution (1-min) SYMH index. For an accurate investigation of ionospheric electric field, one needs to remove the effect of these magnetospheric currents from the magnetic data. For that Choudhary et al. (2011) and Yamazaki and Maute (2016) have proposed a least square technique that is, by fitting, at a given station, the linear trends to nighttime data of the H component and SYMH index. The fitting equation can be written as where T is the time in Julian days. Equation 2 is for five hours nighttime data of each day n. The coefficients C 1 , C 2 , and C 3 are determined by the method as described below. Consider the equation where n H contains the H data during the nighttime, that is, with k denoting the total number of nighttime data points and n A is a matrix contains the T and SYMH index, as follows The symbol M in Equation 3 depicts a vector which contain the fitting coefficient, namely After calculating n H and n T , the fitting coefficients can be determined using the least square estimation technique (Yamazaki & Maute, 2016). For that one can write where t denotes the transpose matrix operation. Once the coefficients C 1 , C 2 , and C 3 are known, the corrected variation in H component can be found by the relation The daily quiet variation (Sq) is calculated from H  over five quiet days ( 5 j  ), that is, Finally, we can determine the desired ionospheric disturbance current (D iono ) by using Equations 8 and 9 in Equation 1.
The magnetic disturbance D iono during a storm can be associated with two phenomena, namely DP2 and D dyn . Here the former is a short period magnetic disturbance associated with PPEF (Nishida et al., 1966) and is simultaneously observed at all longitudes during the main phase of a magnetic storm (Nishida, 1968). Whereas the D dyn is associated with DDEF (Bulusu et al., 2018;Fathy et al., 2014;Le Huy & Amory-Mazaudier, 2005) and results in the anti-Sq oscillations, which are often observed during the recovery phase. Moreover, the D iono has also contribution from other sources such as partial ring current during the early phase of a storm and sometime Sq-like oscillations during the recovery phase, which cannot be associated D dyn (Younas et al., 2020). Hence, simply applying filters to D iono for the estimation of D dyn may not provide an accurate information. Here, we propose a wavelet-based method, namely semblance analysis (Cooper & Cowan, 2008) for the extraction of anti-Sq oscillation. Semblance analysis compares the local phase relation between two data sets as function of time and wavelength. This can be done by performing a cross wavelet transform (CWT) of two time series (Torrence & Compo, 1998) which gives amplitude (α) and the local where Im and Re denotes the imaginary and real parts, respectively. The semblance is then evaluated as (Cooper & Cowan, 2008)   Semblance cos , where n is a positive odd integer. From above equation it is clear that the semblance ranges from −1(anti-correlated) to 1 (positive correlated).
Finally, D dyn can be computed by comparing Sq variation of a station with corresponding D iono value, followed by the extraction of only those periods which are anti-correlated with Sq oscillation, that is, Here, the function semblance(Sq, D iono ) determines the correlation between two signals. Since we are interested in anti-Sq signatures, thus the contributions having a positive correlation with Sq are neglected. The factor α in Equation 11 denotes a cross-wavelet power amplitude, which quantifies the strength of anti-correlated periods. The semblance analysis provides a phase correlation between two signals which ranges from −1 (anti-correlated) to 1 (positive correlated). However, it does not provide the strength of such correlation (or anti-correlation), for example, signals A and B can have semblance of −1, similarly A and C may also have the same semblance, but they can have very different strengths. Such difference of strength in two signals is quantified by the factor α, as determined by CWT. Thus, the semblance is further multiplied by α to find the relative strength of anti-correlated periods (see Equation 9 of Cooper & Cowan, 2008). Duration of D dyn is estimated by considering the time interval from the start of disturbance till the semblance reduces to a threshold value 0.3, after which there is a weak anti-correlation and cannot be related to the signatures of D dyn . In principle, such weak signals remain non-zero for relatively longer duration.

Joule Heating of Upper Atmosphere
Here J x , J y , and J z are ionospheric currents flowing along East, North, and upward directions, respectively.  The solar wind speed increases from 420 to 750 km/s while AE index shows a major auroral activity after the CME with a maximum peak (2,690 nT) observed at 1650 UT.

Case 1
In Figure 3b, we present the same analysis as we did for Figure

Case 4
Figure 5 corresponds to the space weather event of April 4-10, 2010. This case has a weak CME at the beginning with streams of HSSWs lasting till April 10, 2010. In Figure 5a, we present the same analysis as in Figure 2a but for the period April 4-10, 2010. The arrival of a weak CME is marked by dotted vertical line on April 5, 2010 at 0740 UT. This CME is associated with an increase in IMF (4-21 nT), IMF Bz (3-18 nT) and the solar wind speed reaches 500 to 810 km/s. Then, IMF Bz starts oscillating till April 10, 2010 which is a key indicator of the HSSW streams. The AE index remained high (1,000 nT>) till April 10, 2010. In Figure 5b the anti-Sq oscillation in D iono is observed at all three stations for the whole considered period, which is also confirmed in D dyn plots, generated by semblance analysis, in Figure 5c. The duration of D dyn is different at all three longitudes, for example, the longest D dyn disturbance is observed at Africa with the duration of 4.5 days followed by America and Asia lasting, respectively for 4.2 and 3.9 days.

Comparison Among the Five Cases
In Sections 3.1-3.5, we have described the geomagnetic activity during the five space weather events (one from each category). In this regard cases 1, 2, and 3 correspond to CMEs with D dyn observed, respectively at three, two, and one sectors, respectively. Whereas the other two cases, namely 4 and 5, describe the space weather events of CME + HSSWs and HSSWs, respectively. In this section, we look at the possible reasons to explain the observed trends in D dyn during these events.
During the case of March 31, 2001, CME strikes the Earth at 0050 UT. The observed D dyn is generated as result of the Joule heating at high latitudes. Figure  During the magnetic storm of September 2001, CME commenced on 25th at 1900 UT. Here we note that there is a short section of JH having a maximum value of 1600 GW at 2230 UT as depicted in Figure 7c. This weak JH could not generate disturbance dynamo at global level and is confined to only a particular sector (as in Figure 3c).
For the case 4, there is a HSSWs during the recovery phase as indicated by high solar wind speed and oscillatory IMF B z . The arrival of CME is indicated by a vertical line in Figure 4   this event there are two main episodes of JH of upper atmosphere as shown in Figure 7d. The first (second) peak occurred at 0900 (1305) UT having a maximum value of 990 (018) GW. Apart from these two peaks JH remains relatively high till the mid of April 08, 2010, which is due to the presence of HSSWs during the recovery phase, which allows energy input for long durations. This JH has enhanced the duration of D dyn at three sectors as observed in Figure 4c, that is, the disturbance lasts for 3.9, 4.5, and 4.2 days, respectively.
The last scenario that is, case 5 corresponds to a purely HSSWs event, where the AE index depicts multiple peaks of energy inputs from August 24 to August 28, 2010. Figure 7e shows the JH estimation during the HSSWs this event. Both JH and AE indicate that there is a large energy input from August 23 to August 29, 2010, however magnitude of JH in this case is smaller as compared to the CME events discussed earlier. The D dyn started first at Asian sector followed by African and American counterparts. The temporal duration of observed D dyn is maximum at Asia and subsequently Africa and America, respectively. Analysis of these five typical storms shows that during the CME events disturbance lasts for 1-2 days, whereas in case of HSSWs, we see that duration of D dyn may last for more than 04 days. The oscillatory IMF Bz for long time, during the HSSWs events, allow long period of Joule heating in the auroral zone. Furthermore, whenever there is HSSWs along with the CME, D dyn disturbance is intensified and lasts for several days after the storm.

Generalization to 19 Storms
We have computed D dyn for 19 different space weather events. These considered storms, depending on the sources (CME/HSSW/CME + HSSW/Several CMEs), are of several categories. Moreover, the CME events are further categorized with D dyn observed in 1, 2, or 3 sectors. In Table 2 we have grouped the cases according to the different categories of events (CMEs, CMEs + HSSW's, HSSWs, Several CMEs). Table 3, based on the data of Guam, M'Bour and Kourou, presents the results of our analysis for 19 selected events. In the same table columns 1 to 6 show, respectively the type of storm, SSC time in UT, time of maximum CP with duration, duration of MP with duration, AE peaks with time in UT and duration of recovery phase. Columns 7 and 8 present the strength and duration of D dyn at Asian, African, and American sectors, respectively and highest value of D dyn during each case is highlighted with bold text. Figure 8a depicts the maximum duration of the observed D dyn for each storm (from left to right) corresponding to CME only, CME + HSSWs and HSSWs events, orderly. Here we note that D dyn duration is largest (from 19 selected events) for the storm 06 (an HSSWs event). However, it can be inferred from the same figure that duration of D dyn for HSSWs and CME + HSSWs events is relatively long as compared to CME (only) events. The mean and standard deviation are 2.25 days and 0.81 for CME, 4.30 days and 0.72 for CME + HSSW's and 4.30 days and 1.69 for HSSWs. Figure 8b presents the maximum strength of D dyn for each storm (from left to right: CME only, CME + HSSWs, and HSSWs only, respectively). The strength of D dyn at a station is estimated by following Le Huy and Amory Mazaudier (2008), namely from minimum of D iono when it is anti-phase to S q . Prior to calculating the minimum of D iono , a moving average filter is applied by following the methodology of Fathy et al., 2014    is generally high for CME events, for example, it is large values for storms 5 and 15. The mean strength and standard deviation are −113.52 and 86.16 nT for CME, −83.16 and 14.68 nT for CME + HSSWs, and −50 and 26.27 nT for HSSWs. Analysis of Table 3 leads to the following comments.
1. For most of the considered cases, that is, 14 out of 19, the D dyn disturbance is observed in the three sectors. 2. When there is an HSSW (alone or with a CME) D dyn is observed in the three sectors of longitude, in all the events and the duration of the D dyn disturbance is longer. 3. We also note that for the cases of HSSW the peaks of AE are weaker (AE < 1,000 nT) than for CME instances (AE from 1,200 to 2,006 nT), however they last for long periods and hence are more effective in producing D dyn .

Comparison With Le Huy and Amory-Mazaudier (2005)
Le Huy and Amory-Mazaudier have analyzed six magnetic storms (shown in gray in Table 3), by using three (different) magnetic observatories (Huancayo in America, Adis Ababa in Africa and Bac Lieu in Asia). While comparing our findings with the said study, we note that three of the considered storms, namely 5, 9, and 17 depict the same trends as reported by Le Huy & Amory-Mazaudier, 2005. However, for the events 10, 14, and 16 there are major differences in Asian and American sectors. Thus, we need to analyze these storms for all the available observatories (Guam, Bac Lieu, Adis Ababa, Mbour, Kourou, and Huancayo) as follows.  Table 3) for which Le Huy and Amory-Mazaudier observed the D dyn signatures at Asian and African sectors. Figure 9a shows the signatures of D dyn at all longitudinal sectors. Moreover, the stations BCL and GUA-belonging to the same longitudinal sector, namely Asia-have a different response. This might be due to the fact that Sq varies, as caused by the planetary waves, on a daily basis and can affect the evaluation of D dyn magnetic signatures. At the station AAE we observe long, as compared to MBO, D dyn signature. This is probably due to the fact that AAE lies within EEJ region and has contributions from other physical processes such as enhanced Cowling conductivity (Cowling, 1932). In the equatorial region-due to the fact that the Earth's magnetic field is almost horizontal-the conductivity is increased and is called the Cowling's conductivity. Grodgi et al., 2017 described the relations between the various electrodynamic parameters at the equator. In contrast to Le Huy and Amory- Mazaudier (2005) we have observed signature of D dyn in the American sector (HUA), this can be due to the use of a different observatory, namely KOU in the same sector. Figure 9b shows the D dyn (from top to bottom) at GUA, MBO, AAE, KOU, and HUA during the storm 14 of Table 3. Here one notes a different signature of D dyn as compared to Le Huy and Amory- Mazaudier (2005). There are two possible reasons; first we have employed a different method to extract D dyn and second, the use of different observatories in the two studies. Moreover, Le Huy and Amory-Mazaudier have considered the anti-Sq signature only on the day after the storm. Fejer et al., in 1983 studied the signatures of disturbance dynamo electric field over Jicamarca in F-region during a magnetic quiet day just after a storm, when there was no prompt penetration of the magnetospheric convection electric field. In the said analysis, they have found perturbed electric fields around 16-24 h after the onset of the storm. In a similar study Le Huy and Amory-Mazaudier considered the ionospheric disturbance dynamo from magnetic data, during a magnetic quiet day and found the anti-Sq oscillations during the recovery phase. Fathy et al. (2014) studied the disturbance dynamo D dyn during the coronal hole event of April 2005 and separated the D dyn perturbation due to the ionospheric disturbance dynamo (Blanc & Richmond, 1980) from DP2 perturbation (Nishida, 1968) which is caused by penetration of the magnetospheric convection (Vasyliunas, 1970). on ionospheric disturbance current to estimate the D dyn perturbation and found that the main disturbance associated with D dyn has a periodicity of around 20-28 h. Younas et al. (2020) recently considered the asymmetric disturbances in the magnetic storm of August 2018. In this work we have proposed a new approach for the estimation of ionospheric disturbance current and analyzed 19 storms due to different solar perturbations, that is, CME, CME + HSSW and HSSW.

Discussion
Our study reveals several interesting features, which are summarized as following.

Effect of Onset Time (UT) and Strength of Storm for Detection of D dyn During CME Generated Storms
Figures 8a and 8b (left) shows duration and strength of D dyn during the CME generated storms. Here one notes that, in contrast to HSSWs generated activity, the disturbance dynamo effects are not always observed at all longitudes. Accordingly, we have divided the CME generated storms in three categories as depicted in Table 2. In Figures 2-4, we have presented storms from each category. Figure 7 describes the Joule heating estimation and the corresponding AE index. During the CME generated storm-depending on its strength-IMF Bz turns southward for long durations and thus allowing large energy transfer into the magnetosphere. This behavior is quite different from HSSWs generated storm in which IMF Bz remains oscillatory. In our case the Joule heating estimates for the three selected storms have different patterns. For case 3, there is a sharp pulse of Joule heating, which quickly drops down to the normal values, while for cases 1 and 2 Joule heating remains for much longer intervals. Hence, for the third case, the said sharp pulse could not generate the dynamo effect globally while during the first two cases the large energy input for long time periods generates thermospheric winds globally. Consequently, the disturbance dynamo becomes strong enough and can be detected at all three longitudes. Hence, during the CME generated storm duration of IMF Bz is crucial in generating D dyn globally.

Longitudinal Variation of HSSW-Generated Storms
We have analyzed two HSSW events (storms 06 and 19), in this regard Figure 8a shows that, from all the 19 events, storm 06 has the longest duration of D dyn . However, the strength of HSSWs is not comparable to the CME generated counterparts (Figure 8b). Both storms have caused D dyn effect at all longitudes, hence interestingly the HSSW-generated storms have long duration of disturbance dynamo. It looks that oscillatory IMF B z frequency plays a key role in the observed signatures of HSSWs generated storms. As indicated by SYM-H index, the HSSWs events could not generate strong ring currents. Furthermore, during the HSSWs events IMF B z remains oscillatory from the start till the end of a magnetic storm, which allows energy to be transferred from solar wind to the magnetosphere in short impulses and lasting for several days. Thus, strong ring current could not develop during these events. During a CME-in comparison with HSSWsthere is a sudden increase in auroral activity, which lasts for a short duration and hence may not produce disturbance dynamo globally. Whereas the continues energy input to auroral ionosphere, during the HSS-Ws, effectively heats the polar region and resulting in the signatures of DDEF at all longitudes (Rodriguez Zulaga et al., 2016).

Comparison of Results With Le Huy and Amory-Mazaudier (2005)
We have also presented the comparison of our findings with an earlier study by Le Huy and Amory-Mazaudier (2005) as depicted in Table 3 (gray). For the storms 5, 9, and 17 we have found similar trends, however for the storms 10, 14, and 16 our findings are quite different. This disagreement is due to the different methodology and observatories used by Le Huy and Amory- Mazaudier (2005), who have selected the events for which there was no auroral activity on the day after a magnetic storm. That means that the associated disturbance DP2, due to the penetration of the electric field, was zero and only the D dyn disturbance-due to the disturbed ionospheric dynamo-was present. Thus, there was no effect of DP2 and no filtering was required, but this method is only valid for the day after a storm in certain special cases. However, D dyn disturbance can also take place on the day of storm (some hour after SSC) and is considered in our study.

Effect of Season on the Duration of Storm
The storms 1, 2, and 3 of Table 3 occurred in different seasons during the same descending phase of solar cycle 24. These storms have almost the same origin, that is, CME with streams of high-speed winds during the recovery phase. Maximum duration of D dyn , as observed from magnetic data, for storms 1, 2, and 3 is 5.2, 3.5, and 04 days, respectively. And the maximum strength of D dyn for these storms is, orderly −96, −95.0, and −42 nT. This shows that the storms occurring in equinox season are most likely to have strong and long-lasting disturbance dynamo, which agrees with the theoretical prediction of Huang (2013). During a magnetic storm large amount of energy deposition, at polar regions, heats the thermosphere and disturbance thermospheric neutrals winds are generated which flow toward the equator (Blanc & Richmond, 1980). During the solstice season, there is a background wind flowing from summer to winter hemisphere which is enhanced during a disturbed period. Consequently, the equatorward generated storm winds penetrate the winter hemisphere, the resulting disturbance currents from each hemisphere are not comparable to each other. However, for the equinox season there is no background wind flowing between the two hemispheres and the disturbed winds in both hemispheres are comparable, which causes a strong change accumulation at the equator and thus resulting in strong and long-lasting disturbance dynamo. . This indicates that for storm 1, energy inputs to ionosphere-thermosphere system, as indicated by AE index, remains high for relativity long duration as compared to storms 2 and 3, which causes a strong and long-lasting disturbance dynamo in the equinox season (storm 1) in comparison with storms occurring in solstice season (storms 2 and 3).

Main Difference in the Response to CME, HSSWs, and CME + HSSWs Generated Storms
The D dyn may be observed in one, two or all sectors depending on the strength, season and starting UT time of a storm. However, we have found that storm with HSSWs followed by CME or only HSSWs have shown the effect of D dyn at all the longitudes, thus making it global. This difference in the response of CME and HSSWs can be associated with oscillatory IMF B z which allows continuous energy input to the magnetosphere-ionosphere system and thus heating of the upper atmosphere. Whereas, during the CME events, IMF Bz turns southward and thus allowing large energy transfer. However, in contrast to HSSWs, such events last for short durations and hence do not generate enough Joule heating to cause a global D dyn. This effect is evident during the weak CME events as illustrated in Table 3.

Conclusion
To conclude, we have analyzed the magnetic signatures of ionospheric disturbance dynamo (D dyn ) with a new method. In this regard the analysis has been performed for 19 different space weather events, for which we have found some interesting features of D dyn which are summarized as follows.
1. The period of D dyn oscillations is found to be between 16 and 28 h, which agrees with some earlier theoretical predictions (Huang, 2013;Rodríguez-Zuluaga et al., 2016). 2. The D dyn during the CME generated storms may be observed, depending on the strength of storm, in one, two or three longitudinal sectors. This fact can be related to southward excursion of IMF B z . The storm with very long period of IMF Bz, or more than one episode of southward directed IMF Bz, may result in a global generation of D dyn , which generates the observed magnetic signatures at all longitudes 3. In contrast to CME generated event, the D dyn during HSSWs or CME + HSSWs is detected at all longitudes. This effect is related with the oscillatory behavior of IMF Bz during the HSSWs, which allows continuous energy to build up in the thermosphere. Furthermore, the HSSWs generated storms are most likely to have long duration of D dyn . 4. Our analysis of three magnetic storms, having almost the same global variations, reveals that the storms occurring in the equinox season are expected to have a strong D dyn as compared to events in other seasons.
5. The magnetic storms with multiple CMEs show multiple episodes of D dyn and can be associated with corresponding episodes of IMF Bz southward. 6. Joule heating analysis from AE index and BATRUS model indicates that the CME, HSSWs and CME + HSSW events have quite different heating patterns of the upper atmosphere. The CME events that have D dyn in only one longitude show sharp and short duration of high energy input. Conversely, the HSSW and CME + HSSW-generated storms have long duration of Joule heating. Probably, this is the reason why HSSW-generated storms tend to have long duration of D dyn .
This study emphasizes that HSSW-driven events usually could not develop strong ring currents which is a proxy to measure the strength of a magnetic storm. However, these events may cause significant perturbation in the ionosphere-thermosphere through penetration of electric fields from high to low latitudes for several days. Hence, HSSWs should be taken into account while modeling and estimating ionosphere response during disturbed periods.