Solar Cycle and Solar Wind Dependence of the Occurrence of Large dB/dt Events at High Latitudes

We investigate sharp changes in magnetic field that can produce Geomagnetically Induced Currents (GICs) which damage pipelines and power grids. We use one‐minute cadence SuperMAG observations to find the occurrence distribution of magnetic field “spikes.” Recent studies have determined recurrence statistics for extreme events and charted the local time distribution of spikes; however, their relation to solar activity and conditions in the solar wind is poorly understood. We study spike occurrence during solar cycles 23 and 24, roughly 1995 to 2020. We find three local time hotspots in occurrence: the pre‐midnight region associated with substorm onsets, the dawn sector often associated with omega band activity, and the pre‐noon sector associated with the Kelvin‐Helmholtz instability (KHI) occurring at the magnetopause. Magnetic field perturbations are mainly North‐South for substorms and KHI, and East‐West for omega bands. Substorm spikes occur at all phases of the solar cycle, but maximize in the declining phase. Omega‐band and KHI spikes are confined to solar maximum and the declining phase. Substorm spikes occur during moderate solar wind driving, omega band spikes during strong driving, and KHI spikes during quiet conditions but with high solar wind speed. We show that the shapes of these distributions do not depend on the magnitude of the spikes, so it appears that our results can be extrapolated to extreme events.

magnetic oscillations in the pre-noon sector produced by the Kelvin-Helmholtz instability (KHI) operating on the flank magnetopause (e.g., Apatenkov et al., 2020;Engebretson et al., 2020;Juusola et al., 2015;Kataoka & Pulkkinen, 2008;Ngwira et al., 2018;Pulkkinen & Kataoka, 2006;Weigel et al., 2002;Weigel et al., 2003). Rogers et al. (2020) recently undertook a study of spikes observed in 1-min cadence SuperMAG data to predict "return levels" for "return periods" between 5 and 500 years. They noted main peaks in occurrence associated with the pre-midnight, substorm region, the cusp, and subsolar region. Schillings et al. (2022) performed a similar study and identified two main hotspots of spikes in the pre-midnight and dawn sectors during selected geomagnetic storms. Zou et al. (2022) studied the occurrence of >100 nT min −1 spikes in SuperMAG data during storm periods in 2015 and 2016 to identify the associated auroral signatures, including poleward expanding auroral bulges, auroral streamers, poleward boundary intensifications, omega bands, and pulsating auroras. In this study we also identify spikes in SuperMAG data, but perform a more longitudinal study over all levels of geomagnetic activity during solar cycles 23 and 24, years 1995-2020. We identify the solar wind conditions that are favorable for spike generation, and study the solar cycle dependence of the spikes. In the main, we focus on spikes with dB/ dt > 300 nT min −1 , as spikes of this magnitude were responsible for power grid blackouts during the 2003 Halloween storm (Kappenman, 2005). The focus on large dB/dt means that our results apply to high latitudes only: more modest values of dB/dt occur at mid and low latitudes (e.g., Rogers et al., 2020;Smith et al., 2019Smith et al., , 2021Smith et al., , 2022 but will not be included in our analysis.

Observations and Discussion
The SuperMAG repository curates data from ground-based magnetometer networks across the globe. We use 1-min cadence measurements of the magnetic field from SuperMAG for the years 1995-2020, inclusive. The SuperMAG analysis technique (Gjerloev, 2012) subtracts the background field from each component, and we refer to the residuals as the N (North-South, positive northwards), E (East-West, positive eastwards), and Z (vertical, positive downwards) perturbations. We then refer to large minute-on-minute changes in these perturbations as ΔN, ΔE, and ΔZ "spikes." If we refer to a spike in any component, we use ΔB. Our method of analysis is similar to previous dB/dt studies using SuperMAG data (e.g., Rogers et al., 2020;Schillings et al., 2022;Zou et al., 2022). We find very few significant spikes at mid and low latitudes, and so focus on geomagnetic latitudes above 50° in the northern and southern hemispheres.
We first investigate the distribution of "Disturbance Polar" magnetic perturbations produced by ionospheric currents at auroral latitudes, which are often referred to as DP1, DP2, and DPY. These are associated, respectively, with substorms, general convection, and cusp dynamics (Mansurov, 1969;Milan et al., 2017;Nishida, 1968;Obayashi, 1967;Svalgaard, 1973). Figure 1 shows the occurrence of perturbations greater than 300 nT in the N, E, and Z components measured by SuperMAG magnetometers, on a grid of 2° in magnetic latitude and 0.5 hr in magnetic local time (MLT). The percentage occurrence of positive (>300 nT) and negative (<-300 nT) perturbations are shown in red and blue, respectively, on a logarithmic scale. The percentage represents the fraction of time in each grid cell that each perturbation is greater than the threshold; this has been normalized to account for the non-uniform latitudinal distribution of SuperMAG stations. Observations from the northern and southern hemispheres are combined, though in the case of the Z component, the polarity of the southern hemisphere measurements is reversed (due to the N axis pointing polewards and equatorwards in the NH and SH, respectively). The 300 nT threshold has been picked somewhat arbitrarily; the shapes of the distributions do not change markedly for other values of the threshold. Observations from three representative years are presented: 2008 being in the depth of a solar minimum, 2016 being a year of moderate solar activity (declining phase of solar cycle 24), and 2003, a year noted for high geomagnetic activity due to being in the declining phase of the very active solar cycle 23. These three years are chosen to emphasize the range in spike occurrence, from minimum to maximum, over the period considered.
The observed distributions are consistent with the known structure of the auroral currents, which in the main comprise the eastward and westward electrojets in the dawn and dusk sectors (producing the DP2 pattern), and the westward substorm electrojet in the pre-midnight sector (the DP1 pattern). The westwards currents result in southward-directed perturbations across the midnight and dawn sectors (between 18 and 09 MLT), while the eastward electrojet produces a much more limited region of northwards-directed perturbations in the dusk sector (14-19 MLT). The auroral upper and lower electrojet indices, AU and AL (Davis & Sugiura, 1966), which monitor northward-and southward-directed perturbations respectively, are controlled by these two MLT regions.
10.1029/2022JA030953 3 of 12 The peak rate of occurrence of N < -300 nT perturbations is 4%, 7%, and 14% in 2008, 2016, and 2003, respectively. The latitude and MLT extents of the positive and negative perturbation regions broaden as solar activity increases; in 2003 perturbations were regularly seen at latitudes as low as 50° magnetic latitude. As discussed by Imber et al. (2013), in 2003 high average solar wind speed combined with high field strength interplanetary magnetic field (IMF) to produce elevated dayside reconnection rates which resulted in almost double the number of substorms as in each of the previous 7 years and pushed the average location of the auroral oval to unusually low latitudes.
Fewer perturbations were seen in the E and Z components. E perturbations tended to be positive at dawn and negative at dusk, though very much overlapping in active years. Z perturbations were negative at dusk; in the dawn sector they were positive at higher latitudes and negative at lower latitudes, consistent with the expected magnetic signatures of a westward electrojet. Both E and Z perturbations were seen near noon in more active years, associated with cusp currents producing DPY signatures. Occurrence distributions of perturbations exceeding ±300 nT in the N, E, and Z components measured by SuperMAG stations, for three representative years with low, moderate, and high solar activity. Positive and negative perturbations are indicated in red and blue, respectively. The distributions are shown on a magnetic latitude and magnetic local time grid, with noon at the top and dawn to the right; gray circles indicate magnetic latitudes of 80°, 70°, 60°, and 50°. A Feldstein oval for K P = 3 (Feldstein & Starkov, 1967;Holzworth & Meng, 1975) is superimposed for reference.

of 12
We now turn to the occurrence of spikes in the magnetic observations. Figure 2 shows the occurrence of spikes in the same 3 years as presented in Figure 1. A spike is defined as two adjacent 1-min data samples in which one or more of the N, E, Z components of the magnetic field change by more than 300 nT, that is, |ΔN| > 300 nT, |ΔE| > 300 nT, or |ΔZ| > 300 nT. Again, the 300 nT threshold is somewhat arbitrary, but we show below that this does not significantly affect the results. The occurrence distributions are normalized to the number of magnetometer measurements made in each grid cell, as in Figure 1. We note that GICs are produced primarily by disturbances in the horizontal component of the magnetic field (e.g., Mac ; however, we include vertical disturbances in our analysis for completeness. We first consider the results from 2016. In the case of all three magnetic components, spikes are observed near auroral latitudes of 64-74°, concentrated in two main local time sectors, which we loosely refer to as pre-midnight (17-02 MLT) and dawn (02-09 MLT); these are the same "hotspots" identified by Schillings et al. (2022) and previous workers. The boundaries of these MLT regions are indicated in the middle panel; we will refer to the 09 to 17 MLT sector as noon. The pre-midnight hotspot is present in all three components, but the dawn hotspot is most evident in the E component. In the N component, the pre-midnight hotspot seems dominated by negative spikes, that is, ΔN < -300 nT. In 2008, the number of spikes is reduced, and the dawn hotspot is almost absent. The pre-midnight spikes are concentrated near high auroral latitudes of 70-74°. Finally, in 2003 many spikes are observed covering latitudes of 54-80°. The MLT distributions are broader, though still concentrated in two hotspots as before.
We now investigate the occurrence in spikes for the years 1995-2020, which encompass solar cycles 23 and 24, shown in Figure 3. Panel (e) shows the monthly (blue) and smoothed (red) sunspot number. Panel (a) shows the number of spikes of different magnitudes observed in each year: |ΔB| greater than 100, 200, 300, and 400 nT min −1 ; these are plotted on a logarithmic scale to show that the variation is the same irrespective of the threshold used. Panel (b) shows the number of spikes, |ΔB| > 300 nT, in each year on a linear scale, but subdivided by MLT region (17-02, 02 to 09, and 09 to 17 MLT, or pre-midnight, dawn, and noon). Panel (c) shows the proportion of spikes in each MLT region for the same 300 nT threshold. Finally, panel (d) shows the yearly variation in the number of SuperMAG stations in the northern and southern hemispheres located above a magnetic latitude of 50°, that is, the number of stations which are able to observe spikes if they occur. This number varies over the 26 years interval considered, in general increasing from around 100 in 1995 and plateauing near 190 after 2008. Panels (f) and (g) show the yearly distributions of solar wind speed and IMF magnitude as box-and-whisker plots. The boxes show the median and upper and lower quartiles, while the whiskers indicate the upper and lower deciles.
A clear solar cycle dependency in the occurrence of spikes is found. Overall a greater number of spikes was observed in the more active solar cycle 23 as opposed to 24 (even though there were more stations operating during cycle 24). Although there are more spikes at solar maximum than at solar minimum, the number peaks strongly in the declining phase of each solar cycle, especially in 2003. Very few spikes are seen in the noon MLT sector, usually less than 10%. Around solar maximum the rest of the spikes are distributed approximately 50% in the pre-midnight sector and 50% at dawn; at solar minimum the split between pre-midnight and dawn spikes is 75%-25%; during the declining phase the split is roughly 60%-40%.
We now investigate the MLT dependence of the spikes in more detail, and identify a third population in addition to the pre-midnight and dawn hotspots. Figure 4 shows the occurrence distribution of spikes in four representative years, 2019, 2013, 1999, and 2003, roughly in order of increasing solar activity. Here we have selected years to emphasize the changing MLT distributions; we have used 2019 instead of 2008 to show the low activity distribution as there were so few spikes in 2008 that it is less clear than 2019, though the two are consistent with each other. Each panel shows the MLT distribution of spikes in each of the components ΔN, ΔE, and ΔZ, normalized to the peak occurrence. In 2019, the pre-midnight hotspot dominates, and in this population most spikes are in the ΔN component. The distribution for 2013 has this same population, but now the dawn hotspot (04-07 MLT) dominates, with ΔE and ΔZ spikes in the majority. In 1999 the same populations are present, but there is now an increase in ΔN spikes at dawn, and the dawn hotspot now extends to 08 MLT. The extension of the dawn hotspot is even more pronounced in 2003, stretching to 10 MLT, and the occurrence of ΔN spikes has increased again and even dominates in the 07 to 10 MLT sector. We conclude that there are three populations of spikes: the pre-midnight hotspot, 10.1029/2022JA030953 6 of 12 dominated by ΔN spikes, the dawn hotspot dominated by ΔE and ΔZ spikes, and an additional population of ΔN spikes in the pre-noon sector. In all years there is a clear dearth of spikes between 10 and 18 MLT, and a similar dearth between 00 and 03 MLT.
We will later show that the pre-midnight hotspot is associated with substorm onsets, the dawn sector hotspot is consistent with the passage of omega bands, and the pre-noon hotspot is associated with field-line oscillations. This interpretation is consistent with previous studies (e.g., Apatenkov et al., 2020;Engebretson et al., 2020;Juusola et al., 2015;Kataoka & Pulkkinen, 2008;Ngwira et al., 2018;Pulkkinen & Kataoka, 2006;Weigel et al., 2002;Weigel et al., 2003). It is interesting that there is a gap between the pre-midnight and dawn hotspots, indicating that although omega bands are a substorm recovery phase phenomenon, they do not propagate from the substorm onset region, but form in the dawn sector before propagating eastwards. Intense auroral activity can occur in the 00 to 03 MLT sector associated with substorms (Forsyth et al., 2020), but it seemingly does not give rise to magnetic field spikes.
We now turn to a consideration of the solar wind and geomagnetic conditions that favor the occurrence of spikes in the different MLT sectors. We form occurrence distributions of spikes as a function of MLT and the simultaneous value of solar wind speed, V SW , north-south component of the IMF, B Z , a proxy for the dayside reconnection rate, Φ D (Milan et al., 2012), the auroral lower electrojet index, AL, and a measure of the intensity of the ring current, Sym-H, as seen in Figure 5. The solar wind properties and geomagnetic indices are taken from the OMNI data set (King & Papitashvili, 2005). The occurrence distributions are formed on a grid with 24 MLT bins along the horizontal axis and 40 bins along the vertical axis. The occurrence distributions are shown by the contours; grid cells with 10 or more spikes are color-coded by the ΔN, ΔE, or ΔZ component that dominates.
The pre-midnight, dawn, and pre-noon populations, in which ΔN, ΔE, and ΔN spikes dominate, respectively, are apparent in all the occurrence distributions. The pre-midnight and dawn populations occur over a broad range of V SW , so we conclude that solar wind speed is not a controlling factor for these spikes. However, the pre-noon population is only observed for high V SW , greater than about 600 km s −1 . The pre-midnight population occurs for relatively modest geomagnetic activity, with the bulk of the distributions occurring for -10 < B Z < 5 nT, Φ D < 100 kV, AL >-1,000 nT, and Sym-H >-100 nT. On the other hand, the dawn population occurs for more active conditions, -30 < B Z < 5 nT, Φ D up to 300 kV with a peak in the distribution near 175 kV, AL down to −2,000 nT, and Sym-H down to −200 nT and below. Finally, the pre-noon population occurs for quiet geomagnetic activity, with B Z near 0 nT on average and Φ D < 50 kV, but as noted before during periods with high solar wind speed, V SW > 600 km s −1 .
We now show three examples of periods with spikes, beginning with Figure 6, encompassing 12 hr on 25 January 2003. The bottom three panels show Sym-H, the IMF B Y and B Z components, and the solar wind speed and density. The top four panels concentrate on the central 4-hr interval, and show N, E, and Z magnetograms from three representative SuperMAG stations, and the AU and AL indices. In the SuperMAG panels vertical lines show occurrences of |ΔB| > 400 nT spikes, with the color indicating the component containing the spike; several spikes can occur in quick succession, so some lines appear thicker, but are multiple spikes. The inset in the Sym-H panel is a magnetic latitude and MLT dial, showing the locations of SuperMAG stations when they see spikes during the 4-hr interval (this includes all stations, not just the three from the top panels). Again, the color indicates the spike component.
In this first example, IMF B Z is varying back and forth between approximately +6 and -6 nT, the solar wind speed is near 700 km s −1 and the density is 2 cm −3 ; Sym-H shows that this is a non-storm interval. AU and 10.1029/2022JA030953 7 of 12 AL show moderate activity throughout the period, with an intensification in AL just before 9 UT. The SuperMAG observations show relatively clear substorm bays in the N component, just before and around the time of the AL intensification. The stations are located in the pre-midnight sector at this time, consistent with the interpretation that these spikes are associated with a relatively intense substorm onset. This example is typical of the spikes seen in the pre-midnight sector.
The second example is from 31 March 2003 and is shown in Figure 7. IMF B Z is near -8 nT throughout most of the interval, and the solar wind speed is near 600 km s −1 . AU and AL show that this is an interval of quite high activity, AL near −1,000 nT, but Sym-H is only −50 nT. The SuperMAG observations show wave-like activity with an amplitude near 500 nT and a period close to 10-15 min in all three components, characteristic of Ps 6 pulsations which are known to be associated with the passage of multiple omega bands over the stations (Rostoker & Barichello, 1980). Spikes greater than 400 nT min −1 are seen in all three components, but mainly E. The stations are located near 04 MLT when most of the spike activity occurs, consistent with the expected location of omega bands (e.g., Apatenkov et al., 2020;Sato et al., 2017). These Ps 6 pulsations are consistently seen when dawn sector spikes are detected in our observations. Finally, the third example is shown in Figure 8, from 31 October 2003. The solar wind speed is in excess of 1,000 km s −1 , and this is a period of mainly northwards IMF with B Z near +20 nT. Sym-H shows that this is the recovery phase of a geomagnetic storm with the peak Sym-H less than -300 nT. AU and AL show moderate activity, but with two substorm-like onsets. However, it is not these substorms that give rise to spikes observed by Super-MAG, but rather semi-continuous wave activity seen mainly in the N and Z components, with periods close to 4 min. The stations are mostly located in the 08 to 09 MLT sector at this time. We conclude that these are field-line oscillations driven by Kelvin-Helmholtz activity on the dawn flank of the magne tosphere, as a consequence of the high solar wind speed. Although the KHI has been invoked before to explain spikes in the pre-noon sector (e.g., Weigel et al., 2003), it is unclear why a similar population of spikes is absent in the post-noon sector, as the KHI is thought to operate equally on both flanks of the magnetosphere.
While the examples shown here are consistent with the driving mechanisms described, and are representative of the local time differences in the spikes found in our study, clearly the situation is complicated. For instance, Zou et al. (2022) examined closely the auroral forms associated with spikes (at least in the substorm region and the early dawn sector when accessible with ground-based optical cameras) and found that many different types of activity could be identified, including substorm onsets, omega bands, poleward boundary intensifications, auroral streamers, and pulsating auroras. It is known that omega bands and pulsating auroras are generally substorm recovery phase phenomena and that they occur together (e.g., Sato et al., 2015), so it will be difficult to disentangle exactly how each is related to dB/dt. However, the periods seen in pulsating auroras is usually of the order of seconds, so we expect that we will not detect them in our 1-min cadence data. On the other hand, if we assume that the disturbances are created by quasi-sinusoidal wavelike perturbations, then we can estimate the wave amplitude, A nT, required to produce a given ΔB nT min −1 for a given period, T min, as A = (ΔB/2π)T. For ΔB = 400 nT min −1 , the 15-min period Ps 6 omega band activity and the 4-min KHI wave activity require wave amplitudes of 255 and 955 nT, respectively. These are approximately consistent with the perturbations observed in Figures 7 and 8, though the Ps 6 wave-forms are steepened so their amplitude does not need to be quite as large as predicted.

Conclusions
We have investigated the occurrence of sharp changes in magnetic field at high latitudes, which could give rise to GICs detrimental to technological systems, as measured by SuperMAG magnetometers for the period 1995-2020. North-south magnetic perturbations showed that significant eastward and westward electrojets were prevalent, especially in years of enhanced solar activity. The eastward electrojet is mainly located between 14 and 19 MLT, and the westward electrojet between 18 and 09 MLT. We then looked for jumps or spikes in the perturbations of the order of several 100s nT min −1 and showed that these occurred in two hotspots, one between 18 and 00 MLT, and the other between 03 and 09 MLT. These have traditionally been interpreted as spikes caused by substorm onsets and due to the passage of omega bands, respectively, and our observations are largely consistent with this. It is curious to note a relative lack of spikes in the 00 to 03 MLT sector, despite this being a frequent site of auroral activity. We then showed the presence of a third hotspot near 09 MLT, which was consistent with field line oscillations driven by KHI activity on the dawn flank of the magnetosphere. Spikes associated with substorms and KHI occur mostly in the north-south component, whereas spikes associated with omega bands are mainly found in the east-west and up-down components.
The occurrence of spikes shows significant variation with solar activity: spikes occur more at solar maximum than at solar minimum, but are most frequent in the declining phase of the solar cycle. Spikes that do occur at solar minimum are mainly associated with substorms, but these also maximize in the declining phase. Omega band and KHI spikes occur at solar maximum and in the declining phase. The occurrence of substorm and KHI spikes peaked sharply in 2003, a year of high geomagnetic activity driven by high average solar wind speed (see Figure 3f). We investigated the occurrence of spikes with greater than 100, 200, 300, and 400 nT min −1 . Naturally, the occurrence of spikes decreased with increasing magnitude. However, the shape of the solar cycle variation for each threshold was similar, indicating that the observed trends can be extrapolated to more intense spikes.
Finally, we studied the solar wind conditions and geomagnetic activity levels when spikes were observed. KHI spikes were observed mainly during fast solar wind, greater than 600 km s −1 , but otherwise generally low geomagnetic activity. Substorm spikes were observed mostly during more active conditions when the magnetosphere was moderately driven, IMF B Z ≈ -5 nT. Omega band spikes occurred under the most active conditions, when the driving was strong, IMF B Z ≈ -10 nT. Although we have shown an association between spikes and solar activity and solar wind conditions, it is clearly the magnetotail plasma sheet that is the ultimate source of the substorm and omega band phenomena, and further work is required to understand why and when the plasma sheet responds to solar cycle and solar wind conditions in the way that it does. Substorm activity occurs at all phases of the solar cycle, and hence substorm spikes are observed in all years. However, substorms are more intense when the magnetosphere accumulates a large amount of open magnetic flux before onset (Milan et al., 2009). In turn, the polar cap maximizes in size during the declining phase of the solar cycle (Imber et al., 2013), due to the development of low latitude coronal holes on the Sun, which result in high solar wind speeds and high IMF magnitudes (Figures 3f and 3g), which drive solar wind-magnetosphere coupling. Hence, substorm spikes occur most frequently at such times. Although omega bands are a substorm-related phenomenon, they apparently are not associated with weak substorms and hence do not produce spikes at solar minimum. Instead, omega band spikes are seen at solar maximum and during the declining phase. KHI spikes are caused by fast solar wind and so are also observed during the declining phase of the solar cycle. Unlike substorms and omega bands, the KHI does not require ongoing magnetopause reconnection, so southward IMF is not a necessary condition for the generation of these spikes, and they can occur during periods of low geomagnetic activity.
We note that high latitude magnetic field spikes which can cause GICs are not confined solely to geomagnetic storm conditions. At mid and low latitudes, sudden commencements are found to be a major factor in producing spikes (e.g., Smith et al., 2019;Smith et al., 2021Smith et al., , 2022, but we do not find a significant population of high latitude spikes near noon. More comparisons of the driving mechanisms of spikes at high and mid-to-low latitudes are required.  Figure 6, but for 31 October 2003. The geomagnetic activity indices shown were sourced from OMNIWeb. ACE solar wind data is used due to a data gap in OMNIWeb.