3.1. Single Weak Event
As shown in the upper panel of
Figure 1, a strong geomagnetic storm started at 1800 UT on 5 May 1988 (shaded area in period II). The value of the geomagnetic Dst index (black line) then rapidly decreased and reached its minimum value (−160 nT) at 1000 UT the next day. Finally, it recovered to quiet time level after 10 May. During the corresponding period, the geomagnetic AE index also increased to 1334 nT. Before the geomagnetic storm (period I in
Figure 1), the value of Dst remained around zero for more than a week, and the AE index was also generally at a quiet level. During this period, δfoF2 for the 27 ionosondes did not indicate any intense ionospheric disturbance. When entering period II, the ionospheric disturbances detected by most of the ionosondes are very strong, because δfoF2 rise above 0.8 for a dozen hours (lower panel of
Figure 1). However, the goal of this paper is not to prove that the ionospheric disturbance derived by SWM can respond to the geomagnetic activity effectively, as this has already been verified [
7,
18,
20]. In order to prove that the weak ionospheric response to an intense geomagnetic activity is a real geophysical phenomenon, not an artifact of the algorithm, MMM will be used as a comparison method for a weak ionospheric response event during a stronger SGS. As shown in
Figure 2, another geomagnetic storm started at 1000 UT on 23 November 1982, and reached its minimum value (−197 nT) at 1700 UT the next day, and the geomagnetic AE index reached 1500 nT. Obviously, this geomagnetic storm event was more intense than the previous one, but the δfoF2 for this period at most ionosondes did not show a strong ionospheric disturbance (<0.6). The two events are compared—the first (
Figure 1) with an isolated storm, and the second (
Figure 2) with a preceding storm, and it can be seen that their ionospheric responses were very different, especially during period I. In the 1988 event, geomagnetic activity was very quiet in period I, and the corresponding ionospheric state was also very quiet. However, in the 1982 event, there was a geomagnetic storm in period I (as indicated by the Dst index), in which ionospheric disturbance was observed.
To investigate whether this difference resulted from the algorithm, we compare the SWM results with those from the MMM. As shown in the bottom panel of
Figure 2, the δfoF2 derived by MMM displays small ionospheric disturbances in most ionosondes during period I, and the geomagnetic storm is not very strong (Dst ≈ −100 nT) in this period. When the stronger geomagnetic storm occurs in period II (Dst ≈ −200 nT), the ionospheric disturbances are no stronger than during period I in most ionosondes. Further comparison of the ionospheric disturbances in
Figure 1 and
Figure 2 shows that the ionospheric disturbance of period II is weaker in 1982, although the geomagnetic storm of period II in 1982 is obviously stronger than that in 1988. The ionospheric disturbance of 1982 responds weakly to the second strong geomagnetic storm, and when the first smaller storm is over, the small ionospheric disturbance persists during almost the whole of period II. Even after period II, the obvious small ionospheric disturbances can still be found in the 1982 event. Clearly, long-lasting ionospheric disturbances are observed during the period 22 November to 4 December 1982. Although the method has been changed from SWM to MMM, a weak ionospheric response to an intense geomagnetic storm is still observed in period II. This suggests that this phenomenon is not an artifact of the method but might reflect the different geomagnetic conditions in period I during 1982 and 1988.
In fact, the aim of the choice of the two examples in 1982 and 1988 is to verify that the weak ionospheric response to super strong geomagnetic activity is real and not very incidental geophysical phenomenon. It seems that this finding of phenomenon is method independent: both MMM and SWM can identify this kind of phenomenon. Furthermore, comparing with the case of 1988, the ionospheric response is weaker but the geomagnetic activity is stronger in the case of 1982, which can be considered as a much weaker ionospheric response.
3.2. Regional Results and Analysis
No research has yet offered direct or powerful statistical proof that larger ionospheric disturbances appear for a stronger geomagnetic storm. The reason is that the effect of a geomagnetic storm on the ionosphere includes multiple factors, such as the variation in chemical reactions, the effect of the disturbance dynamo electric field, and the enhancement of energy inputs (such as joule heating and particle deposition) into the high latitudes [
21,
22,
23]. These factors mean that the ionospheric response to a geomagnetic storm is complicated. Consequently, even in a very strong geomagnetic storm, the ionosphere may show a weak response in many regions.
To identify events that can truly reflect the weak ionospheric response to geomagnetic storms, the chosen geomagnetic storm events should be as strong as possible. The geomagnetic storm can be classified according to the peak Dst. [
24] refers that the storm while −250 nT ≤ peak Dst ≤ −100 nT is defined as intense storm, and that while peak Dst ≤ −250 nT is defined as very intense one. In the study, in order to obtain enough and standard samples of SGS, the storm with peak Dst ≤ −200 nT is defined the event of SGS. A total number of 33 SGS events have been selected from 1959 to 1990. The response of the ionosphere is defined by the following two important parameters: the maximum value of δfoF2 (MVA) and the average value of δfoF2 (AVA) during a super storm. For a weak ionospheric response, the value of MVA should be in a reasonable range during SGS events, so this does not mean a smaller value is necessarily better. Mikhailov et al. [
11] found that ionospheric disturbance (|δfoF2| > 0.2) often appears in geomagnetic quiet time. For SGS events (Dst < −200 nT), if the MVA is less than 0.2 during the storm event, it is difficult to determine whether the ionospheric disturbance is caused by the strong storm. Therefore, we select cases with MVA greater than 0.2. On the other hand, a stronger geomagnetic storm inputs more energy into the Earth’s upper atmosphere. Therefore, if we assume that the relationship between the ionospheric normal response and the geomagnetic storm is linear, the value of the weak response should be smaller than the value calculated by the linear relation.
Based on these ideas, we select six stations from high to low latitudes in the northern and southern hemispheres, and scatterplots of DstM (the minimum of Dst) against MVA during a geomagnetic storm (Dst < −100 nT) are shown in
Figure 3. The linear relation varies greatly among different stations. The Pearson correlation coefficient (CC) between Dst
M and MVA ranged from −0.132 to −0.261, which implies very weak linear relationship between ionospheric response and geomagnetic activity. The events of very weak ionospheric response during strong geomagnetic storms can be identified in
Figure 3. When Dst
M value reaches −200 nT, the corresponds δfoF2 value varies from 0.73 to 1.46 (seen the blue square). The value of blue square is calculated using the linear fit (red lines), while Dst
M equal to −200 nT. If the ionospheric response (MAV) is less than the value of blue square, this means that it is an ionospheric weak response event during the strong storm (Dst
M ≤ −200 nT). As shown in the
Figure 3, the value of blue square is different in different station, which is ranged from 0.73 to 1.46. To correctly obtain the weak response events from all stations, we define the maximum value of |δfoF2| as 0.6. The shaded areas in
Figure 3 correspond to the δfoF2 values 0.2 to 0.6, and the green dots are the selected weak response events; the value of their ionospheric response is far smaller than that given by the linear relation.
The weak response of the ionosphere is defined as having MVR greater than 0.2 but less than 0.6 during the period of a storm. Based on this definition, the weak ionospheric response events for 27 ionosondes are identified through the two methods (MMM and SWM) as shown in
Figure 4. The weak response events (white bars in
Figure 4) occur more frequently in middle latitudes than in the high and low latitudes in both hemispheres, and the two methods give a similar latitudinal distribution.
In the example of
Figure 2, where the previous storm appeared to suppress the response to the following storm, the interval between the storms was very short. This can be considered as a special SGS event that is referred to as “SGS-PRE” in this paper. To explore whether this special SGS event can contribute to the weak ionospheric response in the second storm, 21 SGS-PRE events have been further selected with the additional condition that the time interval between the SGS and its preceding geomagnetic storm (Dst < −40 nT) is no more than seven days. The numbers of ionospheric weak response (derived by SWM and MMM) to SGS-PRE events (black bars in
Figure 4) are also shown in
Figure 4. The weak response to SGS-PRE shows a similar latitudinal trend to the response to SGS.
To explore whether the SGS-PRE can contribute to weak ionospheric response events, the percentages of weak response events that are SGS-PRE events,
, and those without a preceding storm,
, are calculated and shown in
Figure 5. Here, NSGS-PRE is the number of weak ionospheric response events that occur during an SGS-PRE, NSGS is the total number of weak response events, and
Nnon-SGS-PRE is the number of weak responses during a non-SGS-PRE. The percentage of weak response events with SGS-PRE is greatly higher than the percentage without SGS-PRE in most stations (>18) for both methods. The latitudinal dependence can be found in the percentage with SGS-PRE. The foF2 storm effect has a seasonal dependence, and the change of
ratio and cos χ (where χ is the solar zenith angle) are two important factors responsible for the seasonal variation in the ionospheric F2 layer [
25,
26].
To explore the seasonal variation in the ionospheric weak response, the monthly variation in the number of weak ionospheric response events is studied. As shown in the
Figure 6, the number of weak events for 27 ionosondes can be observed in different months. In the SGS events, obvious seasonal dependence can be found in all ionosondes, and the weak response events easily occur in March/April (spring) and September (autumn). Further exploring the SGS-PRE events, the similar seasonal dependence can be also clearly observed. As a result, in the both of storms and preconditioned storms, the weak events occur mainly at the spring and autumn equinox. In March, the number of weak events is same between SGS and SGS-PRE in all stations. However, in the other months, it seems that there are obvious differences between SGS and SGS-PRE in many stations. In general, the number of SGS is larger than that from SGS_PRE. It is a noteworthy that the maximum number of geomagnetic storms is observed precisely during the equinoxes, and it may be small in other moths [
24]. In fact, the events of SGS are also maximum during the equinoxes, and the number of SGS events are very few in other months (not shown here). It may result in the maximum of weak ionospheric response to SGS during the equinoxes.
3.3. Global Results and Analysis
The previous section focused on the regional ionospheric weak response to SGS and SGS-PRE, and we now examine the global ionospheric weak response behavior. The global ionospheric weak response is typically defined by the number of ionosonde stations that show a weak response during a period of SGS. In this paper, the global ionospheric weak response to SGS is specifically defined as more than 75% of the regions (>21 ionosondes) in ionospheric weak response during an SGS. With this definition, SGS events appear mainly during high solar activity (see SGS in
Figure 7). The percentages for solar min and solar max in the SGS events derived by SWM show that the ionospheric weak response events occur more frequently in solar max (71.4%) periods than in solar min (14.3%); similarly, in the result derived by MMM, the percentage for solar max (43.8%) is also much higher than that for solar min (12.5%).
This phenomenon can be observed by both SWM and MMM (lower panel of
Figure 7). However,
Figure 7 shows that the total number of global weak ionospheric response events derived by SWM (seven events) is critically less than the number of the events derived by MMM (16 events). This is due to the fact that the ionospheric disturbance derived by SWM is more sensitive to geomagnetic activity [
7,
18], which is difficult to observe at a single station; thus, the results derived by the two methods are similar in
Figure 2, but the difference becomes obvious in the global case (
Figure 7). In fact, this difference has already been studied by Chen et al. [
18], who used the SWM and MMM to construct a global ionospheric index of response to geomagnetic storms. The difference is not clearly shown in the regional ionospheric response (
Figure 2) but can be seen in the overall ionospheric behavior (
Figure 7). Similarly, the number of weak response events to SGS-PRE with the SWM (five events) is also less than with MMM (eight events). The global weak response with SGS-PRE events as a percentage of all SGS events can be defined as
for SWM and
for MMM. Here, the
NSGS-MMM and
NSGS-SWM represent the number of weak response events derived by MMM and SWM during an SGS, respectively. The
NSGS-PRE-MMM and
NSGS-PRE-SWM are subsets of
NSGS-MMM and
NSGS-SWM, respectively. They denote weak response events derived by MMM and SWM during an SGS-PRE. With these definitions, the number of global weak response with SGS-PRE events as a percentage of all SGS events is 71.4% and 50% through SWM and MMM, respectively. Although the results vary with method, the percentages suggest that a large proportion of the global ionospheric weak events are SGS-PRE events in both methods.
Our results suggest that SGS-PRE should contributes to global ionospheric weak events, possibly via “long-lasting ionospheric disturbances”. As shown in storm I of
Figure 2, a clear ionospheric long-lasting disturbance can be observed over the whole storm event, and even when storm I recovers to the quiet level, ionospheric disturbance can still be found in most stations. When storm II arrives, the ionospheric background may still be in a disturbed state, which can reduce the sensitivity of the ionospheric response to the storm. Therefore, the physical mechanism of long-lasting ionospheric disturbances might play an important role in explaining the ionospheric weak response caused by SGS-PRE. In fact, long-lasting ionospheric disturbance is not a rare phenomenon, and has been extensively studied. Richmond et al. [
27] used a coupled magnetosphere–ionosphere–thermosphere model to verify and study long-lasting disturbances, which had been predicted by Fuller-Rowell et al. [
1]. Yue et al. [
28] also found obvious long-lasting negative ionospheric disturbance during the recovery phase of a super storm. A similar preconditioning effect was also found in a geomagnetic storm event by Lei et al. [
29]. Based on these studies and our results, the reduction of the ratio of atomic oxygen to molecular nitrogen (O/N
2) might play an important role in long-lasting ionospheric disturbances. The O/N
2 ratio is a parameter that strongly determines production and loss in the ionospheric F2 region and the daytime F2 peak density [
30]. Therefore, if an ionospheric long-lasting disturbance is caused by a previous geomagnetic storm and the ionosphere is still in a disturbed state, it will be difficult for another intense storm to further change the disturbed (O/N
2) significantly, which will result in a weak response in ionospheric peak density or foF2.
The poleward meridional wind disturbance, E × B drift disturbance in the low-latitude ionosphere, and the enhancement of energy (i.e., joule heating) input into the high-latitude ionosphere might also be responsible for the ionospheric weak response. This paper focuses on a preliminary statistical study of the ionospheric weak response, and more advanced physical mechanism studies will be carried out using numerical models in future.