Space Weather Disturbances in Non‐Stormy Times: Occurrence of dB/dt Spikes During Three Solar Cycles

Spatio‐temporal variations of ionospheric currents cause rapid magnetic field variations at ground level and Geomagnetically Induced Currents (GICs) that can be harmful for human infrastructure. The risk for large excursions in the magnetic field time derivative, “dB/dt spikes”, is known to be high during geomagnetic storms and substorms. However, less is known about the occurrence of spikes during non‐stormy times. We use data from ground‐based globally covering magnetometers (SuperMAG database) from the years 1985–2021. We investigate the spike occurrence (|dB/dt| > 100 nT/min) as a function of magnetic local time (MLT), magnetic latitude (Mlat), and the solar cycle phases during non‐stormy times (−15 nT ≤ SYM‐H < 0). We sort our data into substorm (AL < 200 nT) intervals (“SUB”) and less active intervals between consecutive substorms (“nonSUB”). We find that spikes commonly occur in both SUBs and nonSUBs during non‐stormy times (3–23 spikes/day), covering 18–12 MLT and 65°–80° Mlat. This also implies a risk for infrastructure damage during non‐stormy times, especially when several spikes occur nearby in space and time, possibly causing infrastructure weathering. We find that spikes are more common in the declining phase of the solar cycle, and that the occurrence of SUB spikes propagates from one midnight to one morning hotspot with ∼10 min in MLT for each minute in universal time (UTC). Finally, we discuss causes for the spikes in terms of spatio‐temporal variations of ionospheric currents.

10.1029/2023JA031804 3 of 28 Archer et al., 2013;Norenius et al., 2021;Wang et al., 2022).In the first case, it is primarily current systems associated with the magnetopause and the ring current that are affected, while in the latter case, ionospheric currents can be directly affected, for example, in the generation of ionospheric auroral features such as traveling convection vortices (TCVs) (Archer et al., 2013) and throat auroras (Han et al., 2015).Moreover, so called Omega bands (auroral signatures shaped as the inverted Greek letter omega) correspond to ionospheric current systems (Amm et al., 2005) and they have been associated with Pc5 and Pi3 pulsations (Heyns et al., 2021;Jorgensen et al., 1999).Omega bands often occur in the morning sector, but sometimes also around midnight (e.g., Apatenkov et al., 2020;Engebretson et al., 2020;Opgenoorth, Oksman, et al., 1983;Partamies et al., 2017) and it has been shown that they can cause significant GICs (Apatenkov et al., 2020;Vokhmyanin et al., 2021).
Note that there are, of course, many different generation mechanisms for ground magnetic disturbances, and potentially for GICs.However, in this article we do not aim to identify all the mechanisms.Instead, we will discuss the causes for dB/dt spikes in terms of spatio-temporal variations in ionospheric currents.
It has previously been shown that dB/dt spikes are more often occurring in two hotspots in magnetic local time: One hotspot before and around local midnight, and one around local morning (Engebretson et al., 2021;Juusola et al., 2015;Kataoka & Pulkkinen, 2008;Schillings et al., 2022;Viljanen et al., 2001;Vorobev et al., 2019;Weigel et al., 2002).In the following they will for simplicity be denoted the "midnight hotspot" and the "morning hotspot", even though the midnight hotspot also include the pre-midnight MLT sector.Recently, it has been shown that there is a spatio-temporal evolution of spikes during major geomagnetic storms: Schillings et al. (2022) used data from major geomagnetic storms since 1980 and showed that the occurrence of the spikes can propagate from the midnight hotspot to the morning hotspot.Such a propagation could sometimes be observed several times for the same storm.
Spikes in the midnight and morning hotspots have generally been attributed to various different phenomena and current systems (Apatenkov et al., 2004;Engebretson et al., 2019Engebretson et al., , 2020Engebretson et al., , 2021;;Schillings et al., 2022;Viljanen et al., 1999;Viljanen & Tanskanen, 2013): Spikes in the midnight hotspot have, for example, been associated with the westward traveling surge, enhancements in the auroral electrojets, the formation of the substorm current wedge, auroral streamers, poleward auroral expansions, and poleward boundary intensifications (PBIs).Spikes in the morning hotspots have been attributed to, for example, Omega bands and various ULF pulsations in the magnetic field, for example, Pc5 pulsations.
Today, there is a growing space weather awareness and coordinated international efforts to protect our increasingly sensitive, and often irreplaceable, technological infrastructure from space weather hazards (e.g., Koskinen et al., 2017).However, there are still large gaps in our understanding of the physics behind many space weather disturbances, and this is a major obstacle when constructing accurate forecast models.For instance, there are still large gaps in the understanding of the physics behind many types of dB/dt spikes.For example, are spikes caused by highly filamented current systems and/or by systems that have rapid time variations and/or by systems which move quickly above the ground?Moreover, since most focus in the past has been on the occurrence of spikes during highly disturbed and stormy times, very little is known about the spike occurrence during non-stormy times, for example, outside geomagnetic storms and between consecutive substorms.Here, we will show that spikes indeed can be common during non-stormy times, and we will discuss the spatio-temporal characteristics of the spike distributions, and relate to possible ionospheric current systems.
The article is organized as follows.Section 2 contains information about the instrumentation, data, and methods.In Section 3, we discuss four example events, and in Section 4, we discuss the results from our statistical investigation by analyzing the spike occurrence as a function of magnetic local time and magnetic latitude (Secs.4.1-4.2), and over three solar cycles (Section. 4.3).In Section 4.4, we discuss possible ionospheric current systems that could be causing the spikes.Our results are summarized and discussed in Section 5.

Instrumentation, Data, and Methods
We use 1-min resolution ground magnetic field data from the SuperMAG collaboration (https://supermag.jhuapl.edu/) which includes data from more than 500 magnetometers all over the globe (Gjerloev, 2012).Data from the years 1985-2021 are used in our investigation, and the magnetic field baselines are removed as explained in Gjerloev (2012).The SuperMAG data are available in the NEZ coordinate system, where N is north, E east, and Z vertically down.In this article we study rapid time variations in the B N and B E components.The time derivatives, dB N /dt and dB E /dt, are computed as the difference between two consecutive magnetic field data points (B n+1 − B n ) divided by the sampling time of 1 min.Consequently, dB/dt will also have 1-min resolution but sampled in the middle of the time interval between B n and B n+1 .By studying the individual magnetic field components B N and B E separately (and not as the full horizontal component as in some other investigations (e.g., Viljanen & Tanskanen, 2013)), we will be able to separate effects from variations in predominantly east-west current systems (e.g., the eastward and westward electrojets) from variations in current systems with significant north-south contributions (e.g., currents associated with the westward traveling surge).See, for example, the discussion in Section 4.4.
The SuperMAG data are delivered by numerous originators, such as geomagnetic observatories, academic institutions, and national agencies.After being delivered, all data are rotated into the joint NEZ coordinate system, and baslines are removed.This means that the SuperMAG data are not identical to the original data.SuperMAG are not in control of details in the original data, for example, the calibration and the orientation of the original coordinate system, but great care is still continuously taken by the SuperMAG staff to validate and correct the data after delivery.Details in the SuperMAG data processing technique are discussed by (Gjerloev, 2012).Moreover, to make sure that the data are as clean and correct as possible, the authors of this article have manually inspected numerous events, in close collaboration with the SuperMAG PI and the SuperMAG staff.However, errors in the original data, is outside of our control, and they can, of course, be a source of error in our analysis.Moreover, some magnetometers can be more sensitive to detecting spikes for purely physical reasons, for example, due to variations in the ground conductivity and due to the coast effect (e.g., Fischer, 1979;Gilbert, 2005;Pirjola, 2013).However, it is outside the scope of the present study to take into account the conditions of individual magnetometers, such as their calibration and effects from the ground-conductivity and any nearby sea water.
Large excursions in dB/dt are here denoted "dB/dt spikes" and for their selection we require |dB/dt| > 100 nT/min.Note that we treat each 1-min dB/dt sample individually since we do not merge consecutive spikes, similarly to Schillings et al. (2022).
As mentioned in the introduction, there is a non-trivial relation between dB/dt spikes and GICs.Our choice of threshold for spike detection (100 nT/min) can be motivated by previous investigations, for example, by a recent study of the power grid in Sweden, where a horizontal component dB/dt down to 30 nT/min can cause detectable GICs, reminiscent of grounding faults, and 200-500 nT/min can cause a temporary loss of individual transformers (Rosenqvist et al., 2022).In another recent study relating dB/dt variations to GICs in Austrian infrastructure, it was found that dB/dt of the order of 100 nT/min can be associated with cumulative GICs (over an hour at selected transformer stations) larger than 200 Ah (Bailey et al., 2022).Weygand (2021) listed dB/dt for three well-studied previous geomagnetic storms, and found that peak dB/dt values of ∼60 nT/min can correspond to GIC magnitudes up to 40 A. In a statistical study by Viljanen and Tanskanen (2011), a threshold of 60 nT/min was used, and in an early study by Kappenman (2006), it was argued that magnetic field variations of the order of 100 nT/min may be harmful for human infrastructure.However, even though there are investigations that indicate that spikes of the order of ≳ 100 nT/min may cause some harm, it should be noted that our infrastructure is typically designed to withstand strong GICs for quite long times.For example, Swedish transformers are designed to withstand a 200 A DC current for at least 10 min (Rosenqvist et al., 2022).
In this article we investigate the occurrence of dB/dt spikes, with a specific focus on time intervals outside geomagnetic storms, which we denote as non-stormy times.In addition, we investigate the spike occurrence during substorms interval and during the less disturbed intervals between two consecutive substorms.
A geomagnetic storm is a major disturbance in the magnetosphere, and it can last several days.To identify stormy and non-stormy times, we use the SYM-H index computed at the World Data Center (WDC) for Geomagnetism at Kyoto and downloaded from the OMNI web (https://omniweb.gsfc.nasa.gov/).Like the Disturbed Storm Time index (Dst), SYM-H is designed to measure disturbances in the symmetric ring current, but it has a higher resolution than the Dst index (1 min instead of 1 hr) (Wanliss & Showalter, 2006).A geomagnetic storm is typically considered intense if Dst < − 100 nT, moderate if −100 nT < Dst < − 50 nT, and small if −50 nT < Dst < − 30 nT (Gonzalez et al., 1994;Vorobev et al., 2019).Moreover, so called Sudden Commencements (SC) correspond to positive Dst and SYM-H excursions, and they can be caused by, for example, interplanetary shocks (e.g., Lühr et al., 2009).Here, we use −15 nT ≤ SYM-H < 0 to signify "non-stormy" times.
A substorm corresponds to a magnetospheric disturbance that specifically affects auroral latitudes and that has a much shorter duration than a geomagnetic storm, typically 2-4 hr (Partamies et al., 2011).It should be noted that 10.1029/2023JA031804 5 of 28 substorms can occur both inside and outside geomagnetic storms.A substorm is related to the explosive release of energy in the magnetotail, causing ionospheric currents and auroras at high latitudes.The effects on the ground can be measured with the WDC Kyoto indices AE, AU, and AL (Davis & Sugiura, 1966).In this investigation we study dB/dt variations with the largest possible coverage in MLT and Mlat.Therefore, instead of the Kyoto indices AE, AU, and AL we use the corresponding SuperMAG SML, SMU, and SML indices that have a much better global coverage (Newell & Gjerloev, 2011).
There is no rigorous definition of a substorm (Rostoker et al., 1980;T. I. Pulkkinen et al., 2014).Here, we use a method similar to T. I. Pulkkinen et al. (2014) who used AL < − 200 nT for identifying substorm activities, but we use the higher spatial resolution SML: Intervals where SML < −200 nT are here considered as substorms.We allow all intervals with SML < −200 nT to be selected even if they are short, but we smooth SML with a 30 min running average to remove the shortest spikes and weakest events.After selection, substorms that are less than 30 min apart are finally merged into a single event which in the following will be denoted "SUB".
The interval between two consecutive SUBs will be denoted "nonSUB" in the following, and it consequently corresponds to a less disturbed time interval where the smoothed SML ≥ -200 nT.
To determine the geomagnetic disturbance level during the SUB and nonSUB intervals, we compute average values of the disturbance indices: the average SYM-H is computed as the median value during the intervals, and the average SML as the lower 25th percentile.The reason for choosing the median and the 25th percentile, respectively, is due to the different time scales of substorms and geomagnetic storms: SML varies significantly within the SUB intervals and a substorm is classified as intervals where SML shows negative excursions, SML < −200 nT.SYM-H, on the other hand, typically varies on time scales much longer than substorms.
To select non-stormy SUBs and nonSUBs, we compute the median SYM-H and select those intervals where the median value is within the range −15 nT ≤ SYM-H < 0. Computing the median is an efficient way of classifying a large set of intervals from many years of data, but it cannot resolve the SYM-H variations within the individual intervals.For example, we cannot separate intervals with a quasi-constant and zero SYM-H from intervals that possess both positive and negative SYM-H excursions, but where the median is zero.It is outside the scope of the present investigation to study the more detailed behavior of SYM-H, but in future investigations, more advanced SYM-H methods could be developed.
It is well-known that the solar wind and interplanetary magnetic field (IMF) influences the geomagnetic activity on the ground (e.g., Maggiolo et al., 2017).Therefore, as a reference for our ground-based data, we also use OMNI solar wind and interplanetary magnetic field (IMF) data shifted to the nominal bow shock.From 1 January 1995, these are 1-min resolution data collected from IMP8, ACE, and Wind, while prior to this, only 1-hr resolution IMP8 data are available.However, the IMP8 data are very sparse as compared to later years.Therefore, we will only use OMNI data from 1995 and onwards, and the result in.

Example Events
Below we discuss four typical non-stormy (−15 nT ≤ SYM-H < 0) example events during which dB/dt spikes are observed by several ground-based magnetometers.The events are presented in Figures 1-3 and Table 1.We have selected the events to correspond to various conditions: The events are from four different years, from different magnetometer stations, from both SUB and nonSUB intervals, and they include data from various magnetic local times (MLT), and from both the northern and southern hemisphere.The events are denoted SUB I-II and nonSUB I-II.

SUB I
SUB I was observed 2012-03-01 at 09:27:30-14:11:30 UTC corresponding to 21:18-02:05 MLT.It is presented in the first row of Figure 1.Panels 1(a)-(b) show the magnetic B N (red) and B E (blue) components, and the corresponding time derivatives, as seen by the BRW station (Barrow, Alaska) during the entire SUB interval.In total, BRW observed 28 spikes during SUB I, and they are indicated with circles (Figure 1a-1b).The spike threshold (±100 nT/min) is shown with the two dashes horizontal lines (Figure 1b).
The geomagnetic activity context of this event is presented in Figure 1c where SYM-H (red) and SML (blue) are shown within a 12-hr interval surrounding the center time of the SUB (11:49:30).SUB I is highlighted in yellow at the middle of the figure and the three black neighboring SUBs are highlighted cyan and they cover: 05:13:30-06:59:30 (only partly displayed), 14:49:30-16:13:30, and 16:43:30-02:48:30 (only partly displayed).Preceding SUB I, we see a positive SYM-H excursion (07:20-09:30 UTC) with a peak of ∼24 nT around 08:45 UTC.Otherwise, SYM-H is weakly negative during SUB I (median value of about −11 nT) and the interval is therefore considered non-stormy.The average SML index strength for SUB I is ∼ − 480 nT and it dips to ∼ − 700 nT around 10:05 UTC.
As for the solar wind and IMF context (not shown), we find that the solar wind speed is 400-500 km/s and the plasma density is ∼5 cm −3 during approximately a 2-day period preceding the event, but at the time of the positive SYM-H excursion, there is a few hours' enhancement in the solar wind density (average values around 20 cm −3 , peaking at ∼35 cm −3 ).
From Figures 1a and 1b we see that both components, B N and B E , exhibit an approximately synchronous variation with time (a property indicating good data according to Schillings et al. (2022)) which should be caused by overhead ionospheric current systems.We also see that in general B N < 0 and |B N | > |B E |.This implies that the main cause for the ground-based magnetic field observations are spatio-temporal variations in predominantly westward current systems, but with current components also in the north-south direction.This is consistent with the event being a SUB interval observed before and around the magnetic local midnight, that is, when ionospheric current systems such as the westward electrojet, the substorm current wedge, and the westward traveling surge are expected to be significant.We also note that there seems to be four consecutive, but weakening, activations in this westward current system corresponding to the four consecutive depressions in B N (deepest dips at about 10:00, 10:38, 11:30, and 12:17 UTC in Figure 1a).Similar activations are also seen in SML (Figure 1c).
Note that this event includes some relatively strong spikes with magnitudes of 200-300 nT/min, in both positive and negative dB N /dt and dB E /dt, even though SYM-H was weak (∼-11 nT), and the event therefore is classified as non-stormy.The effects from individual spikes can, of course, not be straightforwardly compared since GICs are the results of many different effects, such as details of the ground-conductivity, of the infrastructure itself, and of the time-history of the geomagnetic disturbances.However, it is still interesting to note that the power grid blackout that occurred in Canada during the 1989 strong storm (Dst ≈ −590 nT) had a spike intensity that was observed to be ∼480 nT/min (Boteler, 2019;Kappenman, 2006), which is only about a factor ∼2 larger.
In total 8 stations registered spikes during SUB I (Table 1).The other stations that observed spikes before or near local magnetic midnight (C07, FYU, INK, MCQ, PBK, data not shown) observed similar magnetic field variations as BRW: quasi-synchronous variations in B N and B E and with B N < 0. Most of them observed |B N | > |B E | except for PBK (Pebek, Canada) where |B N |∼|B E |.We suggest that these spikes before and near local magnetic midnight are caused by spatiotemporal variations of predominantly westward currents, but with smaller components in the north-south direction.In addition, the GIM and SMI stations observed spikes in the morning MLT sector.These spikes were associated with a localized disturbance in B N and B E around 13:00 UTC (data not shown).
Figure 2a shows the MLT location of all spikes in SUB I, observed by the eight stations (Table 1).Spikes in dB N /dt (dB E /dt) are shown in red (blue).Note that we do not separate between positive and negative spikes.We have zoomed in on the MLT interval where all spikes are located.Most spikes are located in the evening and midnight MLT sectors.Many spikes are accumulated in sequences, lying on straight lines with slope 1 min/min in UTC-MLT space, with the spike occurrence appearing at later MLT with increasing UTC.For example, there is such a sequence of 28 spikes between 09:40 and 12:45 UTC (21:33-00:37 MLT).These spikes are only observed by one station (BRW), and the observed MLT propagation is just an effect of the station moving to later MLT as Earth rotates, and the spikes occurring continuously above BRW.However, there is also a general trend of a more rapid MLT propagation of the spike occurrence during SUB I, with spike observations starting around 21 MLT (around 09:30-10:00 UTC) and ending around 06 MLT (∼13:00 UTC).This corresponds to UTC-MLT slope of   1. Red and blue correspond to dB N /dt and dB E /dt spikes, respectively, but we do not separate between positive and negative spikes.Note the different UTC and MLT scales.The entire SUB and nonSUB intervals are displayed along the horizontal axis.Note that some spikes may be observed very close to each other in space or time.Therefore, all spikes in Table 1 may not be clearly separated here.Note.The columns list the stations' IAGA identifier, their geographic latitude and longitude, as well as the sum N S of the number of dB N /dt and dB E /dt spikes observed by each station (the MLT interval in which the spikes were observed are in brackets).The magnetic longitude and latitude are not stated since these values are updated every 5 5 years with the new International Geomagnetic Reference Field (IGRF) models.Empty lines in any of the last four columns imply that those stations did not observe any spikes during those events.(2015-12-12 and 2008-03-20) ∼3 min/min, that is, a downward propagation of ∼3 min in MLT for each minute in UTC.Since this MLT propagation trend is observed by several stations (in total 8 stations at different locations, Table 1), we suggest that it corresponds to a real spatial propagation of the ionospheric currents, or a propagation of the activations within the current systems.
Figure 3a shows the Mlat location of all spikes in SUB I observed by the 8 stations.The spikes are observed over an Mlat interval 64°-72°, and the spikes observed by BRW can be seen at Mlat∼70.6°.Note that there is no clear trend of any Mlat propagation of the spikes, for example, as may be expected from the auroral electrojet spreading over larger Mlat ranges during substorms.However, this observation does not exclude any spatial motion of the ionospheric currents.Instead, it only indicates that the spatio-temporal variations, and hence the spikes, were quite evenly spread over 64-72° Mlat.

SUB II
The next example event, SUB II, was observed 2019-05-02 07:52:30-10:45:30 UTC and it is presented in the second row of Figure 1.SYM-H was weakly negative during the event with values between −15 nT and −10 nT (Figure 1f).SYM-H dropped to ∼ − 37 nT about 4 hours before the event, but there are no positive excursions in SYM-H within the entire plotted time interval.As for the previous example event, SML shows some sub-structuring, and there is an SML spike (−720 nT) at about 08:10 UTC.There are no significant solar wind and IMF disturbances related to our event (data not shown).
Figure 1d shows the magnetic field data from the PG4 station (Antarctica).After ∼09:18 UTC we see quasi-periodic ULF ∼3 min pulsations in B N and B E in the morning sector (after ∼06:54 MLT).During the same time, also three other stations (PG5 in Antarctica and SKT and STF on Greenland) observed spikes associated with similar ∼3 min pulsations in the morning sector (data not shown).
ULF pulsations in the post-midnight, morning, and pre-noon sectors have previously been attributed to, for example, Pc5 pulsations (e.g., A. Pulkkinen & Kataoka, 2006;Kataoka & Pulkkinen, 2008;Ngwira et al., 2018), Omega bands (e.g., Apatenkov et al., 2020;Engebretson et al., 2020;Partamies et al., 2017), Pi3 pulsations associated with omega bands (Heyns et al., 2021), or the Kelvin-Helmholtz instability (Kronberg et al., 2021;Weigel & Baker, 2003).ULF pulsations such as Pc5 (a period of a few minutes) and Ps6 (subclass of Pi3 pulsations, tens of minutes) have shown to be possible drivers for GICs (Heyns et al., 2021).Most of these previous studies have focused on pulsations observed during the main phase, or recovery, of large geomagnetic storms.However, it should be emphasized that SUB II was observed during non-stormy times.More detailed studies, preferably also using optical data, are needed to determine the exact cause for the observed ULF pulsations, but this is outside the scope of the present study since our main focus is the statistical investigation of the spike occurrence.
The other stations in  1a).These variations are consistent with spatio-temporal variations in predominantly westward current systems.
Inspecting Figure 2b, we see that here are several spike sequences with an approximate UTC-MLT slope of ∼1 min/min (i.e., caused by the Earth's rotation).The sequences are observed by one station, by nearby stations, or by stations approximately at the same latitude in both hemispheres (Figure 3b).Therefore, we cannot conclude any propagation in the overhead ionospheric systems for these spikes.Instead, we interpret the ∼1 min/min UTC-MLT slope sequences to correspond to pure time variations (or activations) in the current systems above the stations.Moreover, note that close to the beginning of SUB II (08:00-08:24 UTC) there are activations (spikes) over a larger MLT interval (20-01 MLT).
From Figure 3b we see that the spikes at the beginning of SUB II are associated with a more southward and wider Mlat range (65°-71°) as compared to the later spikes (69°-73°), and we conclude that the first set of spikes are correlated with activations in current systems that were observed over quite a large MLT-Mlat region.Note that spikes in the southern hemisphere are indicated with circles in Figure 3b.

nonSUB I
The third example event is a nonSUB interval that was observed 2015-12-12 10:03:30-13:14:30 UTC around local magnetic midnight (22:53-02:17 MLT).The geomagnetic activity context is shown in Figure 1i.Note that the yellow region now corresponds to the present nonSUB example event, while the cyan regions still correspond to the surrounding SUBs.We see that SYM-H was weak with a median of ∼ − 5 nT.SML was also weak, with values typically varying around −90 nT.There are no significant solar wind and IMF disturbances related to this event (data not shown).
The magnetic field and its time derivative observed by the DED station (Alaska) are shown in Figures 1g and 1h.
During most of the event (after ∼10:30 UTC, i.e., after ∼23:40 MLT), B E was close to zero and B N mostly negative.This is expected for the westward electrojet near and after magnetic midnight.However during the first ∼30 min of the event, there were two rapid positive excursions in B N , with B E < 0 and |B N |≫|B E |, hence corresponding to intensifications of eastward ionospheric current systems.Two dB/dt spikes were observed by DED during the northward B N intensifications.Note that SML was weak during this time, and it showed no significant intensifications at this time.Nor were there any intensifications in SML at the end of the preceding substorm that could be associated with the magnetic field intensifications.
In total 8 stations observed spikes during nonSUB I (Table 1).Two of them observed magnetic field variations near local magnetic midnight similar to DED (BRW and JCO, data not shown).Four other stations observed ∼10 min ULF pulsations around local magnetic morning (ATU, CDC, PGC, T53, data not shown).The last station (KTN) observed one spike at 22:25 MLT just at the very end of nonSUB I, that is, just before the next SUB (defined by SML again dropping below −200 nT, see Section. 2 and Figure 1i).
From Figure 2c we see that the spikes first occur around 22:30 MLT at the beginning of the event, and about 2-2.5 hr later in UTC, the spikes occur around 08:30 MLT.This corresponds to a propagation of the spike occurrence of about 4-5 min in MLT for every min in UTC.The spikes of nonSUB I occurs at 70°-74° Mlat (Figure 3c).

nonSUB II
The final example event was observed 2008-03-20 14:09:30-22:06:30 UTC.The average SYM-H and SML during nonSUB II are −4 nT and −64 nT (Figure 1l).Hence, nonSUB II is also observed during non-stormy times.However, note that there is a short SML depression at ∼17:20 UTC (SML ≲ − 200 nT), but it is not considered a SUB in this investigation since the 30-min running average of SML never falls below −200 nT (see Section 2).It can also be noted that this short SML depression was not at all clearly resolved by the original Kyoto AL index (Davis & Sugiura, 1966), but by our SML index which uses a much more global coverage of magnetometers (Section 2).
The four spikes that were observed by the KAV station are shown in Figure 1k.The spikes are observed at 18:20-19:30 UTC corresponding to magnetic local morning (MLT 7:09-8:12) during an interval of ULF pulsations with a period of about 2-3 min.Three other magnetometers in the morning sector also observed similar pulsations to KAV (BRW, CBB, INK, data not shown).
In addition, 10 stations observed spikes associated with damped 5-10 min pulsations in the magnetic local afternoon (MLT 14:22-16:12), starting at ∼17:20 UTC and continuing for about one hour (AMK, ATU, B10, B15, B16, B19, B23, IQA, SKT, STF, data not shown).A quick survey of all available SuperMAG magnetometers (about 300 stations) reveals that a majority of them observed similar pulsations starting at ∼17:20 UTC (data not shown).The pulsations were seen at most magnetic local times, at least in the interval 04-20 MLT, but they were more pronounced in the afternoon sector than at other magnetic local times.The pulsations were strongest in an Mlat interval of about 60°-85°, but there were also similar pulsations at other latitudes, even close to the equator.Note that these damped pulsations coincide with the short SML depression at ∼17:20 UTC in Figure 1l.
Pulsation at the ground, like the ones observed here, could potentially be caused by traveling convection vortices (TCVs) or other vortex phenomena associated with transient changes at the magnetopause such as flux transfer events (FTEs) or solar wind pressure pulses (e.g., Glassmeier et al., 1989;Kim et al., 2015;Shi et al., 2014).
Another possibility is that they could be the effect of the damped ringing of the magnetosphere after the impact of a magnetosheath jet with the magnetopause (e.g., Archer et al., 2013;Dmitriev & Suvorova, 2012;Hietala et al., 2012;Norenius et al., 2021).In a recent statistical investigation, it was shown that ground-based magnetic field signatures similar to our damped pulsations could be caused by magnetosheath jets impacting the magnetopause (Norenius et al., 2021).However, we believe it is unlikely that TCVs or jets could cause such globally occurring pulsations as we observe here.
Moreover, there are no significant IMF or solar wind disturbances in the OMNI data that are associated with the damped pulsations.The solar wind speed was elevated (500-600 km/s) but there were no signatures in the OMNI data of a solar wind sudden impulse or other significant solar wind disturbances that could cause the globally occurring pulsations (data not shown).In addition, the Cluster spacecraft (Escoubet et al., 2001) were upstream of the bow shock in the pristine solar wind prior to the time of the pulsations, but they did not observe any disturbances that could be associated with the damped pulsations (data not shown).
Pulsations on the ground could also be expected from magnetic field-line resonances, but there was no clear change of the pulsation phase with Mlat as would have been expected from a classic field-line resonance (Baddeley et al., 2007;Rae et al., 2007).Similarly, there was no clear phase change with MLT.The pulsations generally had largest amplitudes in the north-south direction.We speculate that the pulsations could be caused by some large-scale ULF wave mode caused by, for example, Kelvin-Helmholz waves on the magnetopause flanks or by magnetotail disturbances.However, there were no spacecraft available for observing any Kelvin-Helmholz waves on the flanks.The THEMIS B and C spacecraft (Angelopoulos, 2008) were in the tail at the time of the pulsations, but according to the THEMIS overview data, they observed no large disturbances.It is outside the scope of the present investigation to study these pulsations further.
According to Figure 2d, there are two groups of spikes: one group in the afternoon sector and one in the morning sector.For the afternoon spikes, we observe a rapid propagation of the spike occurrence toward later MLT.
The propagation is about 3 hr in MLT during 10 min in UTC, or 18 min/min in UTC-MLT space.Moreover, the midnight spikes show a quite wide distribution in Mlat (68°-74° and with some spikes observed in the southern hemisphere, Figure 3d).The other spikes were observed around Mlat 71° except for the three spikes observed at ∼11 MLT observed by CBB at Mlat ∼77°.

Summary of the Example Events
From the example events, we note that spatio-temporal variations in westward currents, sometimes with significant northward or southward components, can cause spikes before and around local magnetic midnight.Moreover, ULF pulsations can cause spikes in the local morning for both SUB and nonSUB events.In addition, there may be other causes for spikes, for example, field-line resonances, Kelvin-Helmholz waves, as well as jets and upstream solar wind pressure pulses and discontinuities interacting with the magnetopause.
Our four example events only constitute a small fraction of all events that are included in our database.Still, they indicate some important features that will be further analyzed in our statistical investigation.One such feature is the MLT propagation of the spike occurrence.An MLT propagation of the spike occurrence in UTC-MLT space of >1 min/min was observed for SUB I (3 min/min) and for both nonSUB events (4-18 min/min).For SUB II we also observed an almost simultaneous outburst of spikes covering a quite large MLT range (20-01 MLT).We conclude that the >1 min/min MLT propagation of the spike occurrence either corresponds to an MLT propagation of the ionospheric current systems, or a propagation of activations within the current systems.
The Mlat distribution of spikes, on the other hand, showed no clear systematic propagation as would, for example, be expected from a southward expansion of the auroral electrojet during active times.Instead, some of the spikes were observed over a rather large Mlat interval, and sometimes also at a similar magnetic latitude in the southern hemisphere.
It is interesting to note that single SUB and nonSUB intervals can contain many spikes, even during non-stormy times.As discussed in Section 2, spikes of the order of ≳ 100 nT/min can be harmful, but human infrastructure is typically designed to withstand strong GICs for quite long times.Hence, we cannot say that the observed spikes could be really detrimental for our infrastructure.However, if several such spikes occur nearby in space and time, there is at least an increased risk for additional (milder) weathering in the infrastructure.

Statistical Investigation
In our database we have collected observations of dB/dt spikes from 1985 to 2021.In Sections 4.1-4.2,we investigate the occurrence of spikes as a function of MLT and Mlat.In Section 4.3 we discuss our statistical results from the perspective of the last three solar cycles.The number of observed events (SUBs and nonSUBs), and the number of available magnetometer stations in each year, are presented in Section 4.3.Finally, in Section 4.4 we discuss some properties of the corresponding ionospheric currents.

UTC-MLT Occurrence
In Figure 4 we show the occurrence of dB/dt spikes as a function of MLT and normalized UTC for seven selected years : 1996, 1999, 2003, 2009, 2012, 2017, and 2021.These years are selected from different phases of the solar cycle (see Section. 4.3).To avoid mixing effects from the northern and southern hemispheres, and since there is a better magnetometer coverage of the northern hemisphere (see Figure S1 in Supporting Information S1), we only include spikes observed in the northern hemisphere.Similarly to the study of Schillings et al. (2022), we separate the occurrence of positive and negative dB N /dt and dB E /dt spikes.In the four columns of the figure we therefore separately display the occurrence of +dB N /dt, −dB N /dt, +dB E /dt, and −dB E /dt spikes.The universal time is normalized so the left (right) half of each panel displays spikes observed during SUB intervals between T 0 and T 1 (nonSUB between T 1 and T 2 ).SUB and nonSUB intervals longer than an upper limit of 8 and 30 hr, respectively, have been removed from the statistics, but it can be noted that only 4% of the SUBs, and 6% of the nonSUBs, are longer than these upper time limits.Black dots corresponds to all spikes (|dB/dt| > 100 nT/min) observed during non-stormy SUB and nonSUB intervals (−15 nT ≤ SYM-H < 0).Cyan and magenta dots highlight those spikes that are observed during intervals with many spikes (N ≥ 20) and few spikes (N < 5) -as mentioned in the introduction, intervals with several nearby spikes are specifically interesting since they may cause additional weathering in human infrastructure.At the bottom of each panel, to the left and right of the vertical dashed line, we display the number of spikes in each interval as well as the number of SUB and nonSUB intervals within brackets.
We see that the number of spikes varies significantly between the years.Fewest spikes, SUBs, and nonSUBs are found in 1999 and 2009, and most in 2003 and 2017 (see also Section 4.3).Both 2003; 2017 hosted famous geomagnetic storms (the Halloween storm in 2003 and the September storm in 2017), and especially 2003 is frequently occurring in previously published storm lists (Schillings et al., 2022;Walach & Grocott, 2019).However, it should be noted that the spikes presented in Figure 4 are only detected during non-stormy intervals, that is, when −15 nT ≤ SYM-H < 0 (Section.2).The number of observed spikes presented in Figure 4 depends on physical processes as well as on the number of available magnetometers.However, as we will show in Section 4.3, the number of available magnetometers in 60° < Mlat <80° (the relevant Mlat, see Section 4.2 below) is approximately constant from 1996 and onwards, so the results in Figure 4 can be compared between the years.
It can be seen that the midnight hotspot (about 20-01 MLT) is activated in both SUB and nonSUB intervals, and for all selected years, even though there are fewer spikes in the midnight hotspot for nonSUBs as compared to SUBs.Moreover, the morning hotspot (about 06-12 MLT) is activated in all nonSUB intervals.In addition, we find that there are very few spikes in the afternoon sector for both SUBs and nonSUBs.These results are also visible for the other years in our database (Figures S2-S6 in Supporting Information S1).Our results that spikes are accumulated in two hotspots are consistent with previous studies (Engebretson et al., 2021;Juusola et al., 2015;Kataoka & Pulkkinen, 2008;Schillings et al., 2022;Viljanen et al., 2001;Vorobev et al., 2019;Weigel et al., 2002).Similarly, the low spike occurrence in the afternoon sector is also consistent with previous studies (Milan et al., 2023).
One interesting observation is that there are SUB and nonSUB intervals with bursts of several spikes (N > 20 spikes, magenta dots in Figure 4) for all the presented years.Hence, even for non-stormy years, there is a risk of consecutive spikes, and even such spikes may not cause detrimental disruptions in human infrastructure, they may cause additional weathering.
Another interesting observation from Figure 4 is a trend within the SUB intervals of the spike occurrence propagating in MLT with increasing UTC from the midnight hotspot toward the morning hotspot.Such an MLT propagation can, for example, be seen for 2017 (for both positive and negative dB N /dt and dB E /dt spikes, Figure 4u-4x), but there is a tendency also for other years and components.The median UTC duration of the SUB intervals included in 2017 in Figures 4u-4x is approximately 81 min.Hence, considering an MLT propagation of the spike occurrence from ∼21 MLT to ∼09 MLT during ∼81 min, this corresponds to an approximate MLT propagation of ∼8.9 min in MLT for every min in UTC.For the other selected years in Figure 4 (1996Figure 4 ( , 1999Figure 4 ( , 2003Figure 4 ( , 2009Figure 4 ( , 2012Figure 4 ( , 2021) ) the MLT propagation is: 9. 4, 8.6, 6.6, 14.4, 10.1, and 10.4   spike observations can propagate quickly from the midnight hotspot to the morning hotspot.We suggest that the cause of this MLT propagation is propagation of the ionospheric currents, or a propagation of activations in the ionospheric currents.Previously, such a propagation toward the morning side has only been observed for major storms (Schillings et al., 2022).Here we show that a similar propagation can also occur during non-stormy times.
It is interesting to compare this MLT propagation during non-stormy times with any possible MLT propagation during more stormy times.In Figure S14 in Supporting Information S1, we present the MLT-UTC spike occurrence for the selected year, but for SYM-H < −30 nT, that is, for times associated with moderate and intense storms (Gonzalez et al., 1994;Vorobev et al., 2019).For the stormy times, we also see an MLT propagation of the spike occurrence from the midnight to the morning hotspot, except for 2009 which was at solar minimum, and the trend was weak for 2012.
In Supporting Information S1 we also show the MLT occurrence of larger spikes during non-stormy times  S11-S14 in Supporting Information S1).As mentioned before, we can compare to the power grid black-out in Canada during the 1989 storm, which happened for a spike intensity of |dB/dt|∼480 nT/min, but during severely disturbed geomagnetic conditions (Boteler, 2019;Kappenman, 2006) when the Kyoto Dst dropped below −590 nT.Moreover, from Supporting Information S1 we see that the midnight hotspot is activated also for these large spikes for most of the years, especially in the SUB intervals.In the nonSUB intervals, we see some activations in either the midnight hotspot or in the morning hotspot, or in both.For some of the years in Supporting Information S1, one can also see an MLT propagation for the larger spikes.
Summing of the number positive and negative spikes in B N and B E for the non-stormy intervals in Figure 4, for the peak year 2017, we find in total 6 ,206 spikes in SUBs and 2, 254 spikes in nonSUBs, or in total 8, 460 spikes in non-stormy intervals, that is, corresponding to ∼23 spikes/day.For a weak year such as 2009, the numbers are 973 spikes in SUBs and 158 spikes in nonSUBs, and hence ∼3 spikes/day.

Mlat-UTC Occurrence
In Figure 5 we show the occurrence of dB/dt spikes as a function of Mlat and normalized UTC during non-stormy times for the selected years.Only observations from the northern hemisphere are included.We see that spikes occur over a quite large Mlat interval, 65° ≲Mlat ≲ 80°, for both SUB and nonSUB intervals.No spikes are observed outside the plotted Mlat range.Moreover, there is no systematic trend of any propagation in Mlat, for example, as would be expected from a southward expansion of the auroral electrojet during active times.Another observation is that many of the magenta dots (N ≥ 20) lie on nearby latitudes and hence can cause some additional weathering in human infrastructure at these latitudes.
The corresponding Mlat occurrence for the years 1987-2021 can be found in Figures S15-S18 in Supporting Information S1.Moreover, in Figure S19 in Supporting Information S1, we show the Mlat occurrence for the selected years for SYM-H < −30 nT.

Overview Over Three Solar Cycles
In Figure 6 we present the spike occurrence over the last three full solar cycles (cycles 22-24), and the top panel shows the number of sunspots (hourly averages).Solar cycle 25 began in December 2019.We see that cycle 22 was the strongest of the three, and 23-24 were continuously weaker, with fewer and fewer sunspots.Solar cycle  1996, 1999, 2003, 2009, 2012, 2017, and 2021 as stated to the right of each row.The columns display positive and negative dB N /dt and dB E /dt spikes.Note that the universal time in each panel is normalized so that the left (right) half of each panel corresponds to SUB (nonSUB) intervals between T 0 and T 1 (T 1 and T 2 ).At the bottom of each panel, to the left and right of the dashed line at T 1 , the number of plotted spikes are presented together with the number of included SUB and nonSUB intervals within brackets.Black dots corresponds to all spikes (|dB/dt| > 100 nT/min) observed during non-stormy SUB and nonSUB intervals (−15 nT ≤ SYM-H < 0).Cyan (magenta) dots highlight spikes observed during intervals with in total <5 (≥20) spikes.

10.1029/2023JA031804
19 of 28 24 is known to be associated with relatively low geomagnetic activity and it is the weakest cycle recorded during space age (Hajra, 2021).The seven selected years that were discussed previously (Figures 4 and 5 Figure 6b shows the number of days per year with IMF B z < 0 nT (black), and with solar wind speed V > 400 km/s (red), ion density N > 7 cm −3 (green), and dynamic pressure P d = Nm p V 2 > 2 nPa (blue), where m p is the proton mass.We see that there is not any significant variations over the years in the number of days with southward IMF and with high ion density.However, the number of days with V > 400 km/s and with P d > 2 nPa vary significantly over the years.The days of fast solar wind had local maxima at the declining phases of solar cycles 23-24, and local minima near the beginning of the inclining phases of cycles 23-25.This is consistent with previous observations showing that the highest solar wind speeds are observed during the declining solar cycle phase when coronal holes extend to lower solar latitudes (Gosling et al., 1976;Mursula et al., 2017Mursula et al., , 2022)).
In Figure 6c we show the number of magnetometer stations included in the SuperMAG database in the northern hemisphere (cyan) and in within 60° < Glat <80° (magenta).We see that the number of stations increased significantly before 1996, and that the number in the northern hemisphere has been quite stable since 1996 (see also Figure S1 in Supporting Information S1).In the following, the number of spikes observed each year will be normalized to the number of available stations within 60° < Glat <80°, that is, the latitude range that is most affected by spikes (Figure 5).Moreover, since there are much fewer magnetometers prior to 1996, we will no be able to draw any strong conclusions about those years.
The next two panels show the number of SUB intervals observed during all SYM-H (Figure 6d), and the number of SUB intervals observed during selected SYM-H ranges (Figure 6e).Note that the black curve is the same in both panels, but that the number of intervals in Figure 6e is shown on a logarithmic scale.In 2007 and later years, there are approximately as many SUBs observed in non-stormy times (−15 nT ≤ SYM-H < 0, solid magenta line) as in more stormy times (SYM-H < − 15 nT, dashed-dotted line).However, before 2007, SUBs in stormy times generally dominate.Times with SYM-H > 0 (dotted) can be associated with gradual or sudden commencements (Hutchinson et al., 2011), and they are considerably less common.Such commencements are typically caused by compressions of the dayside magnetosphere, and they can have a duration of one or several hours.
There is a tendency of the number of SUB intervals decreasing over the years, with the number of SUBs varying around 900 for cycles 22-23 and around 700 for the very weak cycle 24 (Figure 6d).(It should be noted that the number of SUBs is not strongly affected by the small number of magnetometers before 1995, since it is based on the behavior of the SML index, Section 2.) Moreover, the number of SUB intervals varies within each solar cycle, with local maxima at the declining phase (∼1994 for cycle 22, 2003 for cycle 23, and 2015-2017 for cycle 24), and local minima near the solar cycle minima (1986-1987, 1997, 2009, and 2020).This is consistent with the highest solar wind speeds occurring at the declining solar cycle phase (Gosling et al., 1976;Mursula et al., 2017Mursula et al., , 2022)), as well as the formation of corotating interaction regions (CIRs), which together with coronal mass ejections (CMEs), are known to cause increased geomagnetic activity (Richardson et al., 2006;Zhang et al., 2007).Note that the peak in the occurrence of the high-speed solar wind has been observed to shift from the late to the earlier declining phase in solar cycles 23-24 (Mursula et al., 2022).This is consistent with our observations of the peak in the number of SUBs occurring somewhat earlier in the declining phase of cycle 23-24 as compared to cycle 22 (Figure 6d).
The next two panels present the normalized number of −dB N /dt (red) and −dB E /dt (blue) spikes within SUB (Figure 6f) and nonSUB intervals (Figure 6g).Data observed during non-stormy times and during all SYM-H activities (value divided by 10) are displayed with the solid and dashed lines, respectively.Note that the spike statistics are significantly affected by noise before 1996 due to the small number of magnetometers then.From  (i) The ratio between the number of negative and positive spikes in B N (red) and B E (blue) in SUB (h) and nonSUB (i) intervals.In (f)-(i) the dashed lines correspond to all geomagnetic activities and the solid lines to non-stormy times according to SYM-H.Note that the number of spikes during all activities in (f)-(g) is divided by 10.For convenience, the SYM-H unit (nT) is omitted in the legend to the right of the panels.The previously discussed seven selected years (Figures 4 and 5) are indicated in yellow.
1996, on the other hand, there are about 80 stations available within 60° < Glat <80°, and the spike statistics are better.
Like the occurrence of SUB intervals, we find that more spikes are observed in the declining phase of the solar cycle as compared to the other phases.A higher spike occurrence during the declining phase is consistent with recent results from Milan et al. (2023), but here we also show that this is true for both SUB and nonSUB intervals, as well as for non-stormy and stormy times, at least for cycles 23-24 where the statistics are sufficient.For solar cycle 22, this is less clear, even though some tendency can be seen in Figures 6f and 6g.However, as mentioned above, there are much fewer magnetometer prior to 1996, and the data are less reliable.
For both non-stormy times and all activities, we conclude that the spike occurrence, as well as the SUB occurrence, peak in the declining phase of the solar cycle.Moreover, the spike and SUB occurrences reach smallest values near solar minimum and in the beginning of the subsequent cycle.However, the number of spikes varies significantly between the solar cycles.For example, the normalized number of spikes in the declining phase of cycle 23 is 2-3 times higher than in the declining phase of cycle 24.The fact that more spikes were observed during solar cycle 23 is consistent with it being stronger than cycle 24 (Figure 6a).However, the number of normalized spikes does not vary linearly with the number of sunspots, since cycle 25 only recorded ∼1.5 times higher number of sunspots than cycle 24.
Quite expected, we see that there are considerably less spikes during non-stormy times than during all SYM-H activities: For SUB intervals, there is approximately one order of magnitude less spikes during non-stormy times than during all activities (compare solid and dashed lines in Figure 6f).For nonSUB intervals, the difference is smaller, but still there are significantly less spikes during non-stormy times (solid and dashed lines in Figure 6g).
Comparing Figures 6f and 6g, we also see that there are in general fewer spikes during nonSUB intervals than during SUB intervals.However, the number of spikes during non-stormy times is still not negligible for most solar cycle phases.For example, for non-stormy SUB intervals, the normalized number rarely falls below 10 spikes/station (Figure 6f).
Another observation from Figures 6f and 6g is that the number of −dB N /dt spikes is almost always larger than the number of −dB E /dt spikes (compare red and blue lines).This indicates the importance of spatio-temporal variations in east-west aligned ionospheric current systems, over north-south aligned systems, for the occurrence of spikes.
In Figures 6h and 6i we show the ratio between the number of negative and positive spikes in the B N (red) and B E (blue) components, for non-stormy times (solid lines) and all activities (dashed lines), and for SUB (Figure 6h) and nonSUB intervals (Figure 6i).Note that the ratio can be noisy for years with few spikes (e.g., 2009).For both non-stormy times and all activities, as well as for both SUB and nonSUB intervals, we find that there are in general significantly more −dB N /dt spikes than + dB N /dt spikes: after 1995, the B N ratio varies approximately in the interval 1.1-2 for both SUBs and nonSUBs (red lines).The B E ratio, on the other hand (blue), is closer to 1 (approximately in the range 0.9-1.3)for both SUBs and nonSUBs, hence implying a more equal number of −dB E / dt and +dB E /dt spikes, but with a slight excess of −dB E /dt spikes.
We present results for all geomagnetic activities (Figure 7b-7e), for non-stormy times (Figure 7f-7i), for SUB intervals (Figure 7b-c, 7f-g), and for nonSUB intervals (Figure 7d-e, 7h-7i).Note that data gaps (Figure 7g, cyan), as well as some noisiness in the data, are caused by poor statistics.The noise in the data from 1996 and onwards is generally caused by division of small numbers close to zero.For example, for years close to the solar cycle minima, there is less dB/dt activity, and in some MLT bins, there can be very few spikes.The noise prior to 1996 is also strongly affected by the small number of available magnetometers.However, we can conclude that the number of −dB N /dt spikes very often dominates over the number of −dB E /dt spikes (ratio >1), indicating the importance of spatio-temporal variations in east-west aligned ionospheric current systems, over north-south aligned systems, for the occurrence of spikes.The clearest exceptions are SUBs in 03-09 for non-stormy times and all activities where the ratio is ∼1 (black and magenta lines in Figures 7b and 7g), In Figure S20 in Supporting Information S1, we present the ratio between the number of −dB N /dt and +dB N /dt spikes and the ratio between the number of −dB E /dt and +dB E /dt spikes.There are significantly more −dB N /dt spikes than + dB N /dt spikes in the MLT 21-24 and 00-03 bins.For all the other cases, there are a more equal number of negative and positive spikes.

Ionospheric Currents
A larger number of −dB N /dt spikes than −dB E /dt spikes in most MLT bins (Figure 7) indicates that spatio-temporal variations in east-west ionospheric currents are be more important for the spike occurrence than spatio-temporal variations in north-south currents.However, from the spike occurrence alone, we cannot fully analyze the  influence of the ionospheric current systems on the spikes.For that we need additional information on the magnetic field itself (B E and B N ), and not only the time derivative.For SUBs in MLT 18-22 and 22-03 (Figures 8a and 8b), we see that the spike distribution primarily covers large B N < 0. This is consistent with most spikes being observed simultaneously with a strong westward electrojet current, as is expected during auroral activities at these magnetic local times when the substorm current wedge also enhances the westward electrojet.A smaller number of SUB spikes (50 + 17 in Figure 8a) are associated with B N > 0 and hence eastward ionospheric currents.As for nonSUBs, we see a similar trend in MLT 18-22 and 22-03 (Figures 8e and 8f): most spikes are observed for B N < 0 (i.e., an westward electrojet), but the B N magnitude is typically smaller than for SUBs.
In addition, both SUB and nonSUB spikes in  can be observed for both B E > 0 and B E < 0, indicating that there may also be significant southward currents (B E > 0) and northward currents (B E < 0) when spikes are observed.However, spikes are more common in the third quadrant (B E < 0, B N < 0) than in the fourth quadrant (B E > 0, B N < 0), suggesting that currents with a northward component play a more important role for spike occurrence than currents with a southward component.
There is an interesting tendency for the spike distribution exhibiting a tail toward large positive B E for large negative B N (Figure 8b), hence corresponding to a southwestward current component.For some of the years in the Supplementary Information, such a tail can also be observed in MLT bins.One possibility is that there is a strong southward current at the front of the westward traveling surge (Opgenoorth, Pellinen, et al., 1983), but this cannot explain tails observed in the 03-12 MLT bin as for some of the years in the Supplementary Information.We conclude that the reason for the B E tail remains somewhat unclear at this moment.
For SUBs in MLT 03-12 (Figure 8c), the spike distribution also predominantly covers B N < 0, but there is less spread in B E as compared to MLT 18-22 and 22-03.We conclude that SUB spikes in MLT 03-12 also are predominantly associated with the westward electrojet, a primary current system at these magnetic local times, but as compared to MLT 18-22 and 22-03, there are less contributions from northward and southward currents, possibly since there is no westward traveling surge at these magnetic local times.
For nonSUBs in MLT 03-12 (Figure 8g), the spike distribution is more evenly distributed between B N > 0 and B N < 0, while the B E scatter is smaller.Omega bands and various ULF pulsations can be expected in the MLT 03-12 sector.Omega bands are expected in the recovery after substorms, and they should have a quite even spread over positive and negative B N , while ULF pulsation in general would be likely to have an even spread over both positive and negative B N and positive and negative B E .It is hence likely that Omega bands make a major contribution to the distribution in Figure 8g.
In the final MLT sector, MLT 12-18, there are much fewer spikes for both SUBs and nonSUBs (Figures 8d  and 8h).This is consistent with the results of Milan et al. (2023) who also observed very few spikes in the afternoon sector.The shape of the MLT 12-18 spike distribution varies significantly over the years, and this is likely caused by the poor statistics where individual events may have large impact on the result.

Summary and Discussions
The occurrence of dB/dt spikes has previously mostly been investigated for geomagnetically active times, for example, during geomagnetic storms (e.g., Kappenman, 2006;Kataoka & Pulkkinen, 2008;Lakhina & Tsurutani, 2016;Ngwira et al., 2018;Pirjola, 2000;A. Pulkkinen et al., 2008;Schillings et al., 2022;Viljanen & Tanskanen, 2013).Moreover, with the exception of Schillings et al. (2022), previous investigations have typically studied spike occurrence using various local magnetometer networks (e.g., Engebretson et al., 2020;Juusola et al., 2015).In this investigation, we have instead focused on the spike occurrence during specifically non-stormy times, which we define as times of −15 nT ≤ SYM-H < 0.Moreover, we have analyzed several years of data (1985-2021) from a global set of ground-based magnetometers included in the SuperMAG database.We have investigated the spike occurrence (|dB/dt| > 100 nT/min) as a function of magnetic local time (MLT), magnetic latitude (Mlat), and the solar cycle phases.In addition, we have divided the data into "SUB" and "nonSUB" intervals.The SUBs have been selected as intervals when SML < − 200 nT, where SML is the SuperMAG version of the Kyoto AL index (Davis & Sugiura, 1966) but with a much better global resolution.The nonSUBs have been selected as the intervals between two consecutive SUBs.Our main results can be summarized as follows.

Figure 1 .
Figure 1.Four example events of dB/dt spikes observed by four different magnetometer stations during non-stormy times.The first (last) two rows correspond to two different SUB (nonSUB) events.The B N (red) and B E (blue) components are shown in the first column, and the corresponding time derivative in the middle.Every 1-min sample is indicated with a dot, and the selected spikes with circles.The dashed horizontal lines in the middle column show the threshold for spike selection.The IAGA station identifier and station position in magnetic latitude and longitude are shown in the titles in the left column.The right column shows the geomagnetic activity context of the events: SYM-H in red and SML in blue.The cyan highlighted intervals are neighboring SUB intervals while the current SUB (row 1-2) and the current nonSUB (row 3-4) are highlighted in yellow at the center of the figures.The time interval shown in the first and second column exactly matches the yellow interval in the right column.The start date is stated at the bottom left together with the MLT range for the presented magnetometer data.Note the different ranges of the vertical axes.

Figure 2 .
Figure 2. MLT distribution of the dB/dt spikes for example, events SUB I-II (a-b) and nonSUB I-II (c-d).The spikes are observed by the stations listed in Table1.Red and blue correspond to dB N /dt and dB E /dt spikes, respectively, but we do not separate between positive and negative spikes.Note the different UTC and MLT scales.The entire SUB and nonSUB intervals are displayed along the horizontal axis.Note that some spikes may be observed very close to each other in space or time.Therefore, all spikes in Table1may not be clearly separated here.

Figure 3 .
Figure 3. Same as Figure 2 but for the Mlat distribution of spikes.The circles indicate spikes observed in the southern hemisphere at corresponding negative magnetic latitudes.

Figure 4 .
Figure 4. Occurrence of dB/dt spikes as a function of MLT and normalized UTC.The rows present data for seven selected years: 1996, 1999, 2003, 2009, 2012, 2017,  and 2021  as stated to the right of each row.The columns display positive and negative dB N /dt and dB E /dt spikes.Note that the universal time in each panel is normalized so that the left (right) half of each panel corresponds to SUB (nonSUB) intervals between T 0 and T 1 (T 1 and T 2 ).At the bottom of each panel, to the left and right of the dashed line at T 1 , the number of plotted spikes are presented together with the number of included SUB and nonSUB intervals within brackets.Black dots corresponds to all spikes (|dB/dt| > 100 nT/min) observed during non-stormy SUB and nonSUB intervals (−15 nT ≤ SYM-H < 0).Cyan (magenta) dots highlight spikes observed during intervals with in total <5 (≥20) spikes.

Figure 5 .
Figure 5. Same as Figure 4 but for Mlat versus normalized UTC.
) are highlighted in yellow.These years are selected from different phases of cycles the cycles: (a) 1996 near sunspot minimum; (b) 1999 in the inclining phase of cycle 23; (c) 2003 in the declining phase of cycle 23; (d) 2009 just at the beginning of the inclining phase of cycle 24; (e) 2012 near the maximum of cycle 24; (f) 2017 near the end of the declining phase of cycle 24; (g) 2021 at the beginning of the emerging cycle 25.

Figure 6 .
Figure 6.Spike occurrence over the last three full solar cycles.(a) Hourly value of the sunspot number.(b)-(i) Yearly averages of solar wind and ground-based observations: (b) Number of days with IMF B z < 0 nT, and with solar wind speed V > 400 km/s, density N > 7 cm −3 , and dynamic pressure P d > 2 nPa.(c) Number of magnetometers in the northern hemisphere and within 60° < Glat <80°.(d) Number of SUB intervals for all geomagnetic activities.(e) Number of SUB intervals for various SYM-H (the black curve is the same as in panel (d) but in logarithmic scale).(f)-(g) Normalized number of −dB N /dt (red) and −dB E /dt (blue) spikes in SUB (f) and nonSUB (g) intervals.(h)-(i)Theratio between the number of negative and positive spikes in B N (red) and B E (blue) in SUB (h) and nonSUB (i) intervals.In (f)-(i) the dashed lines correspond to all geomagnetic activities and the solid lines to non-stormy times according to SYM-H.Note that the number of spikes during all activities in (f)-(g) is divided by 10.For convenience, the SYM-H unit (nT) is omitted in the legend to the right of the panels.The previously discussed seven selected years (Figures4 and 5) are indicated in yellow.

Figure 7 .
Figure 7. Continuation of Figure 6 presenting the ratio between the number of −dB N /dt and −dB E /dt spikes for various MLT intervals.Panel (a) is the same as Figure 6a.In the panels we show data for all geomagnetic activities (b)-(e), for non-stormy times (f)-(i), for SUB intervals (b-c, f-g), and for nonSUB intervals (d-e, h-i).The data are binned into 3-hr MLT intervals: 18-21 (blue), 21-24 (red), 00-03 (green), 03-06 (black), 06-09 (magenta), and 09-12 (cyan).The previously discussed seven selected years are indicated in yellow.Note that the vertical axis range in panels (b)-(i) is intentionally selected not to include the peaks of the most noisy data.

Figure 8 .
Figure 8. Spike occurrence (±dB N /dt and ± dB E /dt spikes plotted together) as a function of B E and B N .The data are from non-stormy times in 2021.SUB and nonSUB data are shown in the left and right column, respectively.The four rows show data in four MLT intervals: 18-22, 22-03, 03-12, and 12-18.The bins cover all magnetic local times, but they are intentionally defined to have different durations.The number in each quadrant states the number of spikes observed here.

Figure 8
Figure 8 therefore shows the spike occurrence (±dB N /dt and ± dB E /dt spikes plotted together) as a function of B E and B N at the time of the spikes (B E and B N are computed as the average over the two neighboring data points, (B n+ B n+1 )/2, where the dB/dt spikes are sampled at time step n + 1/2, see Section.2).Figure8shows the result for non-stormy times for one of the seven selected years: 2021.The result for the other selected years is rather similar (see Figures S21-S26 in Supporting Information S1).In the left and right column of Figure8, we show data for SUB and nonSUB intervals, respectively.The four rows display data in four[3][4][5][6][7][8][9][10][11][12][12][13][14][15][16][17][18].Note that the bins intentionally have different MLT durations, but together they cover all magnetic local times.The number in each quadrant states the number of spikes observed here.

Table 1 All
Spikes Observed by Magnetometer Stations Stations During the Four Example Events SUB I-II (2012-03-01 and 2019-05-02) and nonSUB I-II

Table 1
observed spikes before or around local magnetic midnight but no pulsations (data not shown).Instead, they observed approximately synchronous B N and B E variations with B N mainly southward, B E varying around zero, and |B N | > |B E | (i.e., similar to the observations in Figure : Figures S7-S10 in Supporting Information S1 include spikes with |dB/dt| > 200 nT/min and Figures S11-S14 in Supporting Information S1 include spikes with |dB/dt| > 300 nT/min (the spikes in Figures S11-S14 in Supporting Information S1 hence constitute a subset of the ones in Figures S7-S10 in Supporting Information S1).For most of the years, we see that there is a non-negligible number of large spikes.Most of the spikes are in the range 200 nT < |dB/dt| ≤ 300 nT, but there are also a few spikes with |dB/dt| > 300 nT (Figure