Impact of magnetic storms on the global TEC distribution

The study is focused on the analysis of Total Electron Content (TEC) variations during six geomagnetic storms of 10 different intensity: from Dstmin = – 46 nT to Dstmin = -223 nT. The values of TEC deviations from its 27-day median value (δTEC) were calculated during the periods of the storms along three meridians: American, Euro-African and AsianAustralian. The following results were obtained. For the majority of the storms almost simultaneous occurrence of δTEC maximums was observed along the Asian-Australian and Euro-African meridians at the beginning of the storm. The transition from weak storm to superstorm (the increase of magnetic activity) almost does not affect the intensity of δTEC 15 maximum. The effect revealed for the American sector during two storms was the movement of the disturbance front from Northern and Southern high latitudes towards the equator with the average velocity of ~ 400 m/s. The seasonal effect was most pronounced at Asian-Australian meridian, less often at Euro-African meridian and was not revealed at American meridian. Sometimes the seasonal effect can penetrate to the opposite hemisphere. The character of averaged δTEC variations for the intense storms was confirmed by GOES satellite data. The behaviour of correlation coefficient (R) between 20 δTEC at three meridians was analyzed for each storm. In general, R>0.5 between δTEC averaged along each meridian. This result is new. The possible reasons for the exceptions (when R < 0.5) were provided: time-shift of δTEC maximum at different latitudes along the American meridian, the complexity of phenomena during the intense storms and discordance in local time of geomagnetic storm beginning at different meridians. Notwithstanding the complex dependence of R on the intensity of magnetic disturbance, in general R decreased with the growth of storm intensity. 25

1.In addition to the Figures presented, I suggest adding some graphs with the TEC data observation during storm time period together with the average of the observations on quiet days with±1 standard deviation.
We added examples of observed and median TEC values (see new Fig.3a and Fig. 3b).Median value serves as a quiet time reference.
2. In each graph from Figures 2 to 5, I suggest that the main and recovery phase of each geomagnetic storm be highlighted.For example, include a yellow and gray rectangle on each graph to represent the main and recovery phase of the sto We marked the main, recovery phases (MP and RP) and the end of the storms (Te) with vertical lines in new Figure 1 (left column), Figure 2, Figure 3, Figure 5, Figure 6.
3. Figure 3: The resolution quality of this Figure is very poor.
The original source-file had a good quality but it was reduced when converting to pdf.
In the new version of the manuscript we change the organization of the figure (Now it is Figure 4): now there are three panels (columns), each of which shows the results for the particular storm.Left plots of each panel display variations in the Northern Hemisphere and right plots-in the Southern Hemisphere.
4. I suggest that it be discussed, clearly, how each phase of the storms (main and recovery phases) affect the ionosphere.The disturbances observed in the ionosphere during the storms were more pronounced during the main phase or recovery phase ???? Does the main and recovery phase affect the ionosphere in the same or different ways depending on the intensity of the storm ???If necessary: a) include a new section to discuss only what was observed in the ionosphere during the main phase; b) subsequently do the same process for recovery phase.We agree with the comment.The issue is discussed in the new version of the manuscript in detail.
We thank the anonymous referee №1 for his or her valuable comments on our paper.We attach a new version of the manuscript to this response.The changes in the text are in blue font.

Responses to Anonymous Referee #2 (in blue)
Interactive comment on "Impact of magnetic storms on the global TEC distribution" by Donat V. Blagoveshchensky et al.

Anonymous Referee #2
Received and published: 30 March 2018 General Comments to the Authors: The article is very interesting reporting significant findings.The results are of high quality and mostly well presented.However, there are some issues that need to be dealt with.These include the moving front, the inspection of actual TEC maps published by Madrigal Database, and the preferable usage of 1-min SYM-H index instead of the hourly Dst index.Based on the actual TEC maps, the description of moving front during the 31 December 2015 storm needs to be corrected.Specific Comments: Page (P) 3 Line (L) 7: "object", "subject" sounds better It was corrected.
P4 L10-15: As storms create sudden ionospheric and TEC changes, it would be better to use actual TEC values provided by the Madrigal Database (http://cedar.openmadrigal.org/)than averaged 2-hourly GIM TEC values for storm studies.They use predictions to fill the data gaps and averaging over 2 hours that smooths out the storm induced sudden TEC variations.
We would like to base our analysis on GIM TEC data.Two arguments can be given in favor to use GIM maps.
-The differences between data from GIM and the Madrigal are not essential for the estimation of global variations considered in this paper.P4 L25: The 1-min SYM-H index provided by the OMNI database (https://cdaweb.gsfc.nasa.gov/cdaweb/istp_public/)and by the World Data Center for Geomagnetism, Kyoto (http://wdc.kugi.kyoto-u.ac.jp/aeasy/index.html)would be better than the 1-hour Dst index because of its higher time resolution.This higher resolution makes research more accurate.For example, the minimum SYM-H for the last storm, superstorm, was -234 nT reached at 2247 UT and not -223 nT at 2300 UT given by the hourly Dst index.
New Figure 1 contains both indices of geomagnetic activity.SYM-H values were also added to new Table 1 to describe the storm with higher resolution.
As it is seen from Fig. 1 and also proved by other works (e.g.Wanliss and Showalter, 2006), there is no large difference between Dst and SYM-H to estimate the disturbance development.We marked the main and recovery phases of the storms with Dst-index.First, it was done for the illustrative purposes as Dst-curve is less rugged and, second, because we use classification of the intensity of the storms based on Dst-index.
The corresponding explanations were added to the text.
In Figure 3, the individual plots are too small and their labels are very hard to read.The original source-file had a good quality but it was reduced when converting to pdf.
In the new version of the manuscript we change the organization of the figure (Now it is Figure 4): now there are three panels (columns), each of which shows the results for the particular storm.Left plots of each panel display variations in the Northern Hemisphere and right plots-in the Southern Hemisphere.
P6 L30: I do not agree with the concept of disturbance front moving towards the equator applied to the 31 December 2015 storm.These TEC maps shown (see attached PDF) are from the Madrigal Database.The left column is for the end of 31 December 2015, the right column is for 3 hours latter.As the TEC maps show in the left column, there was a high TEC region in the American longitude sector and over the Pacific Ocean with large data gaps where GIM fills the gap with predicted values.But in these TEC maps, we can see actual TEC values and they show that these high TECs remained simultaneously at equatorial, low-and mid-latitudes.There was a peak over the magnetic equator, which is possibly the nighttime equatorial peak (or anomaly) implying that the vertical equatorial ExB drift was downward directed and drove a reverse plasma fountain that created this equatorial peak.However, at the same time, there were equally high TECs at mid-latitudes over the ocean.According to the velocity value given by the authors, the travelling time is 3 hours between +/-40 GLAT and the equator.The right column shows the TEC maps 3 hours latter.As the storm progressed, we can see in the American sector the much lower TECs and the peaks of the Equatorial Ionization Anomaly (EIA) indicating that the vertical equatorial ExB drift was upward directed and drove a forward plasma fountain that created this EIA.The moving front section should be re-written and explained better because it is not supported by the actual TEC maps: there were equally high values at mid-and low-latitudes and over the equator (see left column).In terms of moving peaks, these actual TEC maps show that the equatorial peak turned into an EIA, characterized by a northern and a southern crest, as the vertical equatorial ExB drift flipped from downward to upward .So, the peak TEC moved from the equator to both hemispheres' lower latitudes and not from mid-latitude towards the equator as the authors claim.
We thank the reviewer for the detailed comment.We will consider these issues carefully in our future analysis.In the new version of the manuscript we withdrew the section about the front moving as more similar cases are required to prove the results.P11 L25: As suggested, the authors should study the TEC maps of Madrigal Database regarding the moving front and make the necessary corrections.The conclusion was withdrawn.P13 L10: Other data types (Dst/SYM-H, GOES) should be acknowledged as well.It was done.
We thank the anonymous referee №2 for his or her valuable comments on our paper.We attach a new version of the manuscript to this response.The changes in the text are in blue font.

Introduction
The changes in the Earth's geomagnetic field provoked by Space Weather events can cause ionospheric disturbances.The last are very complex phenomena.One of the parameters that help to estimate the ionosphere state change is the vertical Total Electron Content (TEC) that is the quantity of electrons in a column of unit cross section (Davies and Hartmann, 1997;Afraimovich and Perevalova, 2006).Usually, TEC is calculated using phase and code delays of GNSS satellites signals received by dual frequency ground-receivers.The ionosphere is represented by a thin shell of zero thickness at the altitudes of the ionospheric F-region when calculating TEC (Shaer et al., 1995;Komjathy, 1997).Though TEC is an integral characteristics (Electron content from the satellite to the ground), it is assumed that it characterizes the state of Fregion of the ionosphere.This is due to the fact that the main contribution to electron content is provided by the ionospheric F-region.In recent years, TEC has been widely used for ionosphere diagnostics for local regions and on a global scale due to availability of signals in all-time, all-weather conditions around the globe (Panda et al., 2014) and the large coverage of GNSS receivers worldwide in comparison to other ground-based instruments such as ionosonde networks, radars, etc.
Despite a large number of publications dedicated to the disturbed ionospheric state, new data are still interesting to analyze.
In the majority of works data of vertical ionospheric sounding and TEC are used together.However, at present, TEC acts as an independent parameter, in particular to estimate disturbances as, for example, in works (Jakowski et al., 2006;Gulyaeva and Stanislawska, 2008).
The choice of events for the analysis usually varies from several storms, for instance 15 cases during 2006-2007 (Cander and Ciraolo, 2010) or 217 events between 2001 and 2015 (Liu et al., 2017), to the detailed studies of a particular event, as in (Astafyeva et al., 2015).In the present work we study the global ionospheric responses to six geomagnetic storms using TEC data.The storms of different intensity (from weak to severe) were chosen within a short time interval (one-year period).The effects of the storms of different intensity on ionosphere were compared.
A number of works addressed global ionosphere variations during disturbances.One of the possible approaches is to study the behaviour of parameters along different meridians (Mansilla, 2011;Astafyeva et al., 2015).The majority of studies of latitudinal or longitudinal dependences of ionospheric responses are limited to some latitude-longitude region, although there are studies of global density distributions.For example, Zhao et al. (2007) suggested the presence of a longitudinal effect of the ionospheric storm caused by geomagnetic disturbance.Rajesh et al. (2016) showed using GIM that mid-latitude electron density enhancements exhibit significant longitudinal dependence.Longitudinal varieties of the ion total density in the equatorial and mid-low latitudinal topside ionosphere at four local times were studied by (Chen et al., 2015).Latitudinal variations between longitudes 40ºE and 100ºE in the Indian zone were addressed by Bhuyan et al. (2002).Nogueira et al. (2013) examined the four-peaked structure in the observed topside ion density and its manifestation as longitudinal structures in TEC over South America.Dmitriev et al. (2013) performed the longitudinal analysis of the day-side ionospheric storms within the region of equatorial ionization anomaly during recurrent geomagnetic storms.Longitudinal features of electron density distributions were studied in (Klimenko et al., 2015;Klimenko et al., 2016) for minimum solar activity using modeling, GPS and satellite observations.The present study addresses the global longitudinal TEC features not limited by one particular latitude-longitude zone.Three longitude sectors being rather far from each other were chosen for the analysis: along the American meridian (100ºW), along the Euro-African meridian (15ºE) and along the Asian-Australian meridian (115ºE).The effects were studied along these three longitudes within the latitude interval between 60ºN and 60ºS.
The storms considered in the present study were also the subject of several case studies mostly for some particular region.For example, Polekh et al. (2016)  The aim of this work was to reveal the features of TEC variations during the particular geomagnetic storms along three meridians: American, Euro-African and Asian-Australian.The tasks were to: (1) obtain TEC variations along each meridian, (2) find if there is any correlation between these variations, (3) reveal if there is a peculiar character of TEC behaviour during the considered storms if compare to the quiet conditions and how this character depends on the intensity of disturbance and on the meridian itself.
2 Data used for the analysis

Parameters of magnetic storms
Six geomagnetic storms within one-year interval between March 2015 and March 2016 were chosen for the analysis.This period lays on the descending phase of solar activity cycle, not far from its maximum occurred in 2014.The majority of the storms occurred during the winter time in Northern Hemisphere (if categorize March as a winter month) and summer time in Southern Hemisphere.We have chosen the storms of different intensity.Figure 1 (left panels) illustrates Dstindex variations characterizing the disturbances.Vertical lines indicate main phase (MP), recovery phase (RP) and the end of the storm (Te).In some cases sudden storm commencements (SSC) are also indicated.
Table 1 provides the information about each event under analysis.The number assigned to each storm is given in the first column.The same numbers are used to label the panels of Figure 1 (between the left and right columns).The dates of disturbances are given in the second column of Table 1.The time moments of the beginning of the main phase of the storm are given in the third column.Minimal Dst-index values are given in the fourth column.Fifth column shows SYM-H index minimum values to provide the full picture of disturbance.Sixth column shows the time of the beginning of the recovery phase of each storm.The last seventh column presents the time moments of the end of the storm (Te).Here, "e" means end.Main and recovery phases were defined based on Dst variation.Te moment corresponded to the end of the storm when Dst value was about (-10 ÷-15) nT or before the next SSC.The geomagnetic storms are presented in Table 1 from the less intense (first line) to the most intense (sixth line) according to the Dst-index.Gonzalez et al. (1994) introduced storm classification: intense storms are characterized by Dst ≤ -100 nT, moderate storms -by -100 nT ≤ Dst ≤ -50 nT, weak storms -by -50 nT ≤ Dst ≤ -30 nT.According to this classification, the storm #1 (14.12.2015) is weak, the storm #2 (06.03.2016) is moderate, the storms #3, #4, #5 and #6 are intense.The last storm (17.03.2015) is called a superstorm in literature because it was the most intense storm of solar cycle 24.Thus, all six considered storms are of different intensities.

TEC data
TEC values were obtained from Global Ionospheric Maps (GIM) produced by International GNSS Service (IGS).
GIM TEC are independently computed by four Analysis Centers of the International GPS Service for Geodynamics (CODE, JPL, UPS, ESA) and then ranked and combined according to the corresponding weight by the International GNSS Service to produce the IGS global vertical TEC maps (Hernandez-Pajares et al., 2009) For each observation point median TEC value was calculated on the basis of 27 previous days for every two hours of the day (UT).Thus, its own median value was obtained for each day every two hours.Furthermore, the deviation of TEC was calculated and plotted during each storm as well as six days before and six days after it following Eq.( 1): where TECobs is the observed value, TECmed27 is a median value calculated for the 27 days prior to the day of observation.

Satellite and geomagnetic data
Data from GOES weather satellites that circle the Earth in a geosynchronous orbit was used in the analysis (https://satdat.ngdc.noaa.gov/sem/goes/).The altitude of their orbit is about 35800 km.GOES-13 is positioned at 75ºW 3 Discussion of results

Specific features of TEC variations during the considered storms
Variations of δTEC were the main source of information about the changes in the ionosphere.According to this data, the bursts of δTEC occurred at the beginning of magnetic disturbance.The duration of these bursts varied within several hours.
The behaviour of δTEC along American, Euro-African and Asian-Australian meridians was studied with 10º step in latitude from 60ºN to 60ºS.

Weak δTEC variations
Sometimes manifestations of disturbance in TEC during geomagnetic storms were weak or absent within the latitude range of ±20º near the equator.Figure 2 provides the example for the storm of December 31, 2015 at the Euro-African sector.
Here, for the economy of space the plots are shown with the 20º latitude step along the longitude.Days in Universal Time (UT) were laid off along the X-axis; additionally markings were laid every 2 hours (UT).To confirm this example the more detailed picture of TEC behavior is considered for the case of latitude 20ºN from Fig. 2. (storm period) in general followed its quiet pattern (Fig. 3 panel a).The maximal TEC deviation from its quiet state reached -28%.Such deviation can be referred to day-to-day variability.In contrast, the different picture was observed for the same latitude 20ºN but at the American meridian.Fig. 3 (panel b) shows the results: the geomagnetic storm first caused the positive and then the negative TEC disturbance with the maximal TEC change of 67% from its quiet state.This particular example proves the presence of weak (almost absent) TEC disturbances within the latitudes ±20º at the particular sector.

Seasonal effect
The presence of seasonal effects in δTEC variations was revealed for the following cases.
(a) During the storm #2 (March 6 th , 2016) the positive phase of disturbance was the dominant effect in δTEC variations during the night hours (UT) between March 6-7 along the Asian-Australian meridian from latitude 60ºN to latitude 0º.In contrast, at the same meridian from 10ºS to 60ºS the positive phase was followed by negative phase.In other words, during this storm the positive disturbance covered the latitudes of winter hemisphere, meanwhile summer hemisphere was characterised by positive disturbance followed by negative disturbance.
(b) Similar picture was observed along the same (Asian-Australian) meridian during the storm #4 (December 20 th , 2015).However, though the general tendency of δTEC was similar along the whole meridian (increase of values followed by decrease), in terms of phases the positive phase followed by decrease of values prevailed in Northern (winter) hemisphere from latitude 60ºN to 30ºN (Fig. 4, panel a).Further, from 20ºN to 60ºS, the δTEC increase followed by the clear negative phase was observed.Here, the "summer" effect penetrated into the "winter" hemisphere.Starting from 10ºN positive phase (sometimes various peaks) was followed by negative phase.At that, the positive phase was in the form of a very intense burst (+ 180% and more) at latitudes between 20ºS and 60ºS.In this case, the "winter" effect penetrated into Northern Hemisphere from South.
To sum up, according to our data (cases (a)-(d)), the seasonal effect consists in general dominance of negative phase (decrease of TEC) in summer and positive phase (increase of TEC) in winter.This conclusion is in accordance with the case study (Kil et al., 2003).In the present study the effect was observed mostly over the Asian-Australian sector and no seasonal effect was registered over the American sector.Kil et al. (2003) addressed the case of magnetic storm of July 20 th , 2000, using GIM and low-orbit satellite data.They revealed clear seasonal effects: a dominance of the negative ionospheric storm in the summer (northern) hemisphere and the pronounced positive ionospheric storm in the winter (southern) hemisphere.Kil et al. (2003) also found that the Northern "summer" negative phase penetrated into the Southern hemisphere.Our results also prove the possibility of penetrating of the seasonal effect to the opposite hemisphere.However, in our case both examples (b) and (d) showed such penetration from Southern to Northern Hemisphere: summer effects and winter effects respectively.Thus, we may conclude that it does not depend on the season itself or on the hemisphere.
The storm analyzed by (Kil et al., 2003) was very intense (Dstmin = -300nT).Our examples prove that the seasonal effect can be observed during the magnetic disturbance of less intensity (but still intense): -98 nT (a), -155 nT (b and c), -204 nT (d).
Here, we briefly mention that Zhao et al. ( 2007) also showed with GIM TEC that during magnetic disturbances a negative phase occurred with higher probability in the summer hemisphere, while a positive phase -in the winter hemisphere.According to these authors, negative phase was most prominent near geomagnetic poles and positive phase was far from polar regions.According to our data within the latitudes ±60°, the positive phase is very probable during the disturbances.At the same time it is not contradictory as each geomagnetic storm is a particular unique event.
To conclude, the seasonal effects had longitudinal dependence: observed mostly over the Asian-Australian sector, sometimes over Euro-African sector and no seasonal effect was registered over the American sector.

Global picture of δTEC variations at three meridians
Figure 5 shows the averaged δTEC behaviour.Each panel (a-f) represents the results for the particular storm: from the weakest (panel a) to the strongest (panel f).Storm dates are indicated below the panels.The time-interval on the X-axis is the interval between the storm beginning and Te (individual for each storm), according to Table 1.Each panel consists of three plots: upper plot represents variations in the American sector, middle plotin Euro-African and the lower plotin Asian-Australian sector.The curve on each plot represents δTEC values averaged along one meridian over the latitudes 60 о N -60 о S with 10 о step (δTECav).In other words, the final δTECav curve represents the average of 13 δTEC values from different latitudes.This averaging is possible because according to our data the tendency of increasing or decreasing of δTEC was the same at different latitudes along one meridian in most cases (without the regard to the phase).The specific cases are described above and also considered below.
First, it is seen that the maximal δTECav lays close to storms main phase beginning.Physically, it is explained by the fact that usually the drastic increase of particle flows from magnetosphere into ionosphere occurs at the beginning of each storm that, in turn, results in TEC disturbance.It is known, that during the development of disturbance the critical frequencies of ionosphere decrease lower than their initial quiet level (Blagoveshchensky, 2011).The same behaviour is observed in TEC: minimum of δTECav values is observed after the increase of δTECav, caused by the main phase of storm.
The main feature seen in the panels "a", "b", "e" is approximately the same time (UT) of δTECav maximum occurrence at all the latitudes along three meridians.In regard to panels "c" and "d", their results were discussed above.To add, the δTECav maximum took place at the same time at Asian-Australian and Euro-African meridians.For American meridian the peaks are shifted in time as it was mentioned before and the peaks themselves are more diffused if compare with Asia and Europe.Let us consider a more detailed picture of each panel of Fig. 5.
Panel (a) has the shortest disturbance duration due to the weakness of geomagnetic storm on December 14 th , 2015.
This weak intensity is the reason of the slow ionospheric response and the particle precipitation occur with a certain delay from storm beginning.At that, the moments of δTEC maximums coincide at three meridians.In panel (b) δTECav maximums were well-pronounced and coincided in time at three meridians during the moderate storm on March 6 th , 2016.storms: the increase of δTECav was followed by its decrease.However, the negative phase was more pronounced if compare with the weak positive phase.
To conclude, there is no dependence of δTECav maximums at three meridians on the intensity of magnetic activity.
We recall that the intensity of storms grows from panel "a" to panel "f", but no increase in δTECav variations is detected.

Global picture of δTEC response to main and recovery phases of the storms.
The ionospheric responses to geomagnetic storms at different observation points have their peculiarities due to the differences in local hours, wind systems, electrical fields and other local effects.In addition, geomagnetic storms in different regions can manifest themselves differently.For instance, the exact moments of geomagnetic storm beginning and its intensity can vary.To estimate qualitatively the global effects of storm phases at different latitudes along three meridians we applied some generalization.First, we used Dst-index as a global measure of geomagnetic field change.Furthermore, as it was mentioned above, the tendencies of TEC increasing/decreasing in most cases were similar at different latitudes along each meridian, thus we can consider the average effects along the meridians.With regard to the phase of TEC disturbance, the picture was similar along each meridian in one hemisphere and sometimes in both hemispheres along the whole meridian.The example of such picture is given in Fig. 6.
The rapid main phase of storm #1 (Table 1) provoked TEC increase (beginning of the positive TEC disturbance) during 3 hours of its duration in both hemispheres at three meridians.The only exception was Euro-African meridian at Southern Hemisphere: TEC was already augmented before the main phase (Fig. 6 panel b).The maximum of TEC bursts at all latitudes and meridians occurred during the few hours after the beginning of the recovery phase.The negative phase followed TEC bursts during the second half of the recovery phase at Euro-African and Asian-Australian meridians in Southern Hemisphere.In Northern Hemisphere and at the American meridian in Southern Hemisphere TEC presented the second positive phase (less intense than the first maximum).
The storm #2 was characterized by rapid Dst decrease (Figure 1) and, consequently, by the short main phase (3 hours as in the previous case).The recovery phase lasted 20 hours.As it is known the ionospheric response to geomagnetic storm can be immediate or with a delay in hours and even days.The last is our case.The effects in the ionosphere were observed during the recovery phase, probably because of the short duration of the main phase.In Northern Hemisphere the positive TEC bursts occurred in the middle of the recovery phase along the whole American and Euro-African meridians as well as in Southern Hemisphere along Asian-Australian meridian with the following decrease of TEC.In Northern Hemisphere at Asian-Australian meridian TEC had more complex behaviour and mostly was increased during the whole period of the storm.
The peculiarities of the storm # 3 were already mentioned in Section 3.1.1(weak δTEC variations within ±20º).
Except this feature, in Southern Hemisphere the main phase of geomagnetic storm caused mostly the positive TEC disturbance and the recovery phase caused one or two negative TEC disturbances.
For the storm #4, it can be assumed that the recovery phase provoked one or two negative phases of TEC disturbance at Asian-Australian meridian.The effects in other sectors were rather different to generalize them.
The case of the storm #5 is more complicated as it developed at the already disturbed background: three SSC provoked by the interplanetary shocks (Astafyeva et al., 2017) of different intensities occurred during the considered interval: 16:46 UT on June 21st, 05:47 and 18:30 UT on June 22nd.An intense geomagnetic storm (storm #5) followed the last SSC with its main phase between June 22 nd and June 23 rd (Fig. 1).In general, along the American, Southern part of Euro-African and Asian-Australian meridians the positive TEC disturbance was observed during the main phase and the negative disturbanceduring the recovery phase.
The superstorm #6 provoked complex effects at different observation points.Among the common features of TEC are the following.Along the American meridian the TEC burst was mostly caused during the main phase in Northern Hemisphere and it was shifted towards the recovery phase in Southern Hemisphere.TEC burst was observed during the main phase along other two meridians.Negative TEC disturbance was detected during the recovery phase at all observation points.
To sum up, the following common features were revealed.During the recovery phases of the weak and moderate storms (#1, #2) TEC reached its maximum globally.Though there are some similar features found, in general the intense storms #3 and #4 provoked rather complex TEC responses without dependence on the phase.During the recovery phases of the most intense storms #5 and #6 negative TEC bays were observed.These results are confirmed with averaged TEC behavior in Fig. 5.It should be mentioned that though some similarities in ionosphere variations during the particular phases of storms were revealed, the whole picture is rather complex.

Data of GОES-13 satellite
To compliment the analysis of Figure 5 and for better understanding of phenomena the results of measurements at GOES satellite were involved in this study.Its orbit in the near Earth space is at the altitude of 35800 km that is in the Earth's magnetosphere.Among the measurements performed at the satellite there were the intensity of X-rays, protons with energies from >1 to >100MeV, electrons with energies from >0.8 to >4 MeV.
GOES data was studied during the periods of all six geomagnetic storms (Fig. 1).The particle flows of protons and electrons were registered for all considered storms.However, for storms #1 -#4 (Fig. 1, Table 1) the intensity of these flows did not differed significantly from its undisturbed rate.Rather high levels of particle flows were observed only for storms #5 and #6.Even for Dst values of order of -150 nT (storm #4) the flows level was rather low and only for Dst being lower than -200 nT it was significant (intense storms #5 and #6 with Dst values being -204 nT and -223 nT respectively).Thus, it was impractical to consider satellite data for the first four storms #1 -#4. Figure 7 shows the flows variations for storms #5 and #6.The moments of storm beginnings (To) and storm ends (Te) are labeled with vertical lines for both storms.Figure 1 shows that the amplitudes and the shapes of Dst curves were close for both disturbances.It was of interest to compare the satellite measurements of high energy particles -protons and electrons.Protons variations (p) are plotted in the upper half of the plots of Fig. 7, electrons variations (e)in the lower parts.The beginnings of the two storms were approximately at the moment of maximal proton radiation and the beginning of minimal electron flows.Then, the decrease of proton flow occurred in the interval To-Te, but electron flows increased from its minimal to maximal values during the same time.In general, the proton and electron flows during magnetic storms are probably not directly connected with electron density in the ionosphere (Afraimovich and Perevalova, 2006).However, it is possible implicitly.The increase in δTECav values (Fig. 5) at the beginning of the storm was probably related to the maximum of proton rates.The decrease in electron flux coincided with δTECav decrease.Further, the drastic growth of electron flux intensity took place leading to δTECav growth in Fig. 5.In particular, for the storm #5 (June 23 rd , 2015) Fig. 5 illustrates δTECav bursts before June 23 rd , then the decrease to the minimum around June 24 th and then again some increase between June 24 th -25 th .Similar picture was observed during storm #6 (March 17 th , 2015): the maximal intensity of the proton flux was accompanied with small δTECav increase (not significant in this case but existing) near the storm beginning (Fig. 5,f) and then the decrease of the flux took place.During March 17 th -18 th the electron flux minimum was observed and then its increase.Thus, the character of δTECav behaviour for two storms in some way is proved by satellite data of energetic protons and electrons.

Similarities and differences of δTEC response at different meridians during the storms
We estimated a degree of correlation between δTECav at different meridians for each storm during the disturbed periods.This interval was different for each storm.Thus, 16 δTECav values were found within To-Te during storm #1; 23 valuesduring storm #2; 25during storm # 3; 49 -during storm # 4; 33during storm #5 and 58during storm #6.The distances in degrees between the meridians are the following: American -Euro-African (Am-E) -115º, Euro-African -Asian-Australian (E-A) -100º, Asian-Australian -American (A-Am) -145º.The shortest distance is between E-A meridians and the largestbetween A-Am meridians.Table 2 shows values of correlation coefficient (R) calculated between δTEC values at different meridians: (1) averaged at along the whole meridian (bold type), (2) averaged along the meridian in Northern Hemisphere (normal type), (3) averaged along the meridian in Southern Hemisphere (italic type).

δTEC averaged along the whole meridians
Table 2 illustrates the following features for averaging along the whole meridian (bold type).
-Rather high degree of correlation (R>0.5)took place between the δTEC variations during storms #1-#5 for all meridians except two values R = 0.148 and R = 0.430 between Asian-Australian and American meridians.This is explained by the time shift of δTEC peak along the American meridian as shown in Fig. 5 (panels c and d).We associate low correlations during storm #6 with the complexness of local phenomena because of the high intensity of the storm (including no correlation in the case A-Am).
-The highest R values (if comparing three pairs of meridians) were found between European and Asian-Australian sectors in five cases of six.
-The highest R values between all three meridians (R>0.5) were during the weakest storm #1.This corresponds to the physics of phenomena.Perturbations and irregularities in the ionosphere are more pronounced during intense disturbances than during moderate or weak disturbances.During the weak storm the ionosphere structure is not significantly changed and its global stability is retained.
-The lowest R values (in bold) took place between Asian-Australian and American sectors if compare to other two pairs at least for five storms of six.It is probably explained by the fact that the distance between the American and Asian meridians is the largest (145 о ).Another possible cause is that storm beginnings were observed in the contrary local time zones (day or night local hours) for these two meridians during all storms under analysis.
This is in accordance with physics of phenomena.However, the transition from the storm #4 to the storm #6 shows inverse dependence: some growth of R instead of its decrease for storm #5.Nevertheless, in general, R behaviour in dependence to the intensity of magnetic disturbance (transition from storm #1 to storm #6) showed the decrease of R values, which is to be expected.The lowest R values were for the most intense storm.

δTEC averaged along meridians in each hemisphere
It is known that TEC behavour has a seasonal dependence (Afraimovich and Perevalova, 2006).As the seasons are opposites in two hemispheres, the effects in North and South can be different.In general, it is revealed that the intense bursts of δTEC took place at subpolar latitudes of both hemispheres.To compare "northern" and "southern" data first the averaging of δTEC was performed along each meridian separately in each hemisphere: between the latitudes 60ºN-10ºN (northern) and then between the latitudes 10ºS -60ºS (southern).Middle and lower panels of Fig. 6 serve the example.Though the averaging along the meridian implies only qualitative, not quantitative estimate of deviations, it was of interest to analyze the effects separately.Table 2 presents the results of R calculations made separately for Northern (normal type) and Southern (italic type) hemispheres.
-For two storms #5 and #6 close by their intensities of disturbance, but different by the season of occurrence (summer/winter and winter/summer) the following is characteristic.R<0.5 in Northern hemisphere (summer) and R>0.5 in Southern hemisphere (winter) at all three meridians during the storm #5.For the storm #6 the opposite picture is seen.R<0.5 in both hemispheres and there was no correlation in cases Am-E and A-Am.But in cases of correlation existence, R was lower in Southern hemisphere (summer) than in Northern hemisphere (winter) when the correlation was detected (E-A).It may be related to the seasonal effect, but more statistics is needed to conclude.
-Comparison of R for Southern and Northern hemispheres shows rather high degree of correlation in both hemispheres simultaneously (R>0.5)only for the weak storm #1.For other storms the number of cases when R<0.5 increases with the disturbance intensity: one case for the storm #2, two cases for the storm #3, three cases for storms #4 and #5, five cases for storm #6.In other words, the difference in R values in Northern and Southern hemispheres grows with the increase of magnetic activity.It results that seasonal effect has impact here.
is confirmed by GOES satellite data of energetic proton and electron fluxes.
5) The analysis of correlation coefficients between averaged δTEC variations at three meridians during each storm showed the following.
-The degree of correlation between averaged along a whole meridian δTEC values at three meridians was rather high (R>0.5).This result is new.There are five exceptions of 18 cases from Table 2:  -The highest coefficients of correlation between averaged along a whole meridian δTEC (all three R>0.5)took place during the weakest storm.This is due to the fact that during the weak storm the ionosphere structure is not significantly changed and its global stability is retained.Comparison of R between δTEC averaged separately in Northern and Southern hemispheres also showed that high degree of correlation for both hemispheres R>0.5 took place only for the weak storm.
The difference between hemispheres increased with the increase of magnetic activity, that probably again is explained by seasonal effect.
-The lowest coefficients of correlation (through all the storms in general) were found between Asian-Australian and American meridians.The reasons may include the largest distance between these meridians and discordance in local time of storm occurrence.
-The not evident, mixed dependence of R on the intensity of magnetic disturbance is common for all three meridians.Nonetheless, the transition from weak to the most intense storm shows the decrease of correlation till the absence or even negative correlations.This result is new.It is confirmed by correlation coefficients between both averaged δTEC and δTEC at each latitude separately.In general, the more the intensity of magnetic disturbance, the lower the correlation rates between δTEC variations at three meridians.
-Calculation of R separately for two hemispheres allowed us to reveal that the most intense δTEC bursts took place at subpolar latitudes of both hemispheres.For two storms 23.06.2015 and 17.03.2015close by the intensity but different by the season the following is revealed.For summer storm 23.06.2015R values were less than 0.5 in Northern hemisphere and more than 0.5in Southern hemisphere between all three meridians.For storm 17.03.2015R values were less than 0.5, but in general, the picture was vice versa: correlation coefficients were lower in Southern hemisphere and higherin Northern (when correlation was detected).The seasonal effect probably plays a main role here.
-For the storm of June 23, 2015, R between δTEC at each latitude for all three pairs of meridians was positive within the latitudes ±60º and ±10º (in both hemispheres) and was rather low or negative within the interval 10Nº-10Sº.
Consequently, the ionosphere processes in equatorial zone were the subject of different physical causes at three meridians.
Table 1.Characteristics of the geomagnetic storms used in the analysis.24 addressed the event of March 17, 2015; Astafyeva et al. (2016) studied ionosphere during June 22, 2015; Chashei et al. (2016) considered ionospheric effects during the storm on December 20, 2015, etc.In our case the focus is on global effects.
longitude and the equator monitoring North and South America and the Atlantic Ocean.GOES-15 is positioned at 135ºW longitude and the equator monitoring North America and the Pacific Ocean.The coverage by two satellites extends approximately from 20ºW longitude to 165ºE longitude.The instruments for near-Earth Space Weather monitoring are installed on board including magnetometer, X-ray sensor, high energy proton and alpha detector, and energetic particles sensor.To estimate geomagnetic conditions, the Dst and SYM-H indices values were used.Both indices are the indicators of global Space Weather effects.Data is freely available by following the link http://wdc.kugi.kyoto-u.ac.jp.Wanliss and   Showalter (2006)  showed that SYM-H index can be used as a de facto high-resolution of Dst-index as they are quite similar to characterize the storms of different intensity.This similarity is also seen in Fig.1(right and left columns).We used Dstindex to define main and recovery phases of the storms: (a) for the illustrative purposes as Dst-curve is less rugged; (b) as we use classification of the intensity of the storms based on Dst-index (Section 2.1).
Fig. 3 (panel a) shows the values of the observed TEC (TECobs, green curve), its 27-day running median (TECmed27, red dotted curve) and standard deviation  for TECobs (blue curve).Main and recovery phases (MP, RP) and the end of the storm (Te) are marked with the vertical lines.Median values serve as a quiet reference.It is seen that TEC observed during December 31, 2015 -January 02, 2016 (c) During the same storm #4 along the Euro-African meridian from December 20 th to December 22 nd (0 UT) the disturbance showed the "positive-negative-positive" sequence of phases from 60ºN to 10ºN.Here, the second positive phase was much more intense and the whole disturbance within the interval 30ºN -0º began earlier.The latitudes of Southern hemisphere 0º-60ºS were covered by the negative phase during December 21 st with preceding positive phase almost disappearing.(d)During the storm #5(June 23, 2015)  along the Euro-African meridian the negative phase in the form of two bays was observed from 60ºN to 0º (Fig.4 panel b).From 10ºS to 60ºS the disturbance had more complex character and included two or more positive phases.At the same time along the Asian-Australian meridian the negative phase was observed between 60ºN and 20ºN (Fig.4 panel c).
Panel (c) illustrates the results for the storm on December 31 st , 2015.Time of δTECav maximums occurrence was the same only at Asian-Australian and Euro-African meridians.Panel (d) illustrates the picture similar to panel "c", but for the storm on December 20 th , 2015.Panel (e) shows the results for the intense storm of June 23 rd , 2015.It was the only storm among the six that occurred during the summer at Northern Hemisphere and during the winter in Southern Hemisphere.However, no specific details were revealed in comparison to other considered storms.Panel (f) shows the results for superstorm of March 17 th , 2015.Though it is the most intense storm among the six, in general δTECav variations do not differ from the other (a) R = 0,148 and R = 0.430, both found between Asian-Australian and American meridians, and (b) low R during the most intense storm #6.Issue (a) is related to the time-shift of δTEC maximum at different latitudes along the American meridian.The reason of the shift is provided.Issue (b) is associated with the complexity of phenomena during the most intense storm.

Figure 1 :
Figure 1: Dst-index (left column) and SYM-H index (right column) variations during the periods of six geomagnetic storms under analysis.Main (MP) and recovery phases (RP) as well as sudden storm commencement (SSC) were marked by vertical lines based on Dst-index variation.

Figure 2 :Figure 3 :
Figure 2: Weak manifestation of TEC effects within the latitudes ±20º during the storm of December 31 st , 2015.MP, RP and Te are marked with vertical lines.

Figure 4 :
Figure 4: δTEC variations for storms: (a) #4 at Asian-Australian meridian; (b) #5 at Euro-African meridian; (c) #5 at Asian-Australian meridian.Left plots of each panel display variations in the Northern and right plots-in the Southern Hemispheres.

Figure 5 :
Figure 5: δTEC averaged along each meridian during the storms.Vertical lines indicate the periods of MP and RP.

Figure 6 :
Figure 6: Results for storm #1: δTEC averaged along the whole meridian (upper panels), along the Northern hemisphere latitudes (middle panels) and along the Southern Hemisphere latitudes (lower panels) at the American (column a), Euro-African (column b) 5 . These final IGS maps were used for this study.
worldwide, thus, it is a useful tool for ionosphere diagnostics on a global scale.