Simultaneous observation of auroral substorm onset in Polar satellite global images and ground-based all-sky images

Polar satellite global images

The Polar satellite ultraviolet imager (UVI) (Torr et al. 1995) provides global imaging of auroras. UVI GIs are captured in the \(\mathrm{N_{2}}\) Lyman–Birge–Hopfield long (LBHL, \(\sim 170\hbox { nm}\)) and short (LBHS, \(\sim 150\hbox { nm}\)) wavelengths and the OI \(\sim 130.4\) and \(\sim 135.6\hbox { nm}\) wavelengths. In particular, the LBHL images monitor the energy flux of precipitating keV-range electrons (e.g., Lummerzheim et al. 1997) and thus can be compared with green-line (557.7 nm) ASIs during substorms. The LBHS images are less useful for this purpose because LBHS emissions are absorbed by the atmosphere en route to the imager, and this absorption depends on the average energy of precipitating auroral electrons. The UVI captures four or five images in each 184-s cycle. The filter and the exposure period usually vary during this 3-min cycle; thus, the practical time resolution is usually 3 min for the same filter and exposure.

In the present study, the UVI captured images with a fixed wavelength (LBHL) and exposure mode at 36.8 s, which enabled a higher practical time resolution of 36.8 s than the usual 3 min. This 37-s resolution is expected to be marginally sufficient for resolving the two stages of the substorm onset sequence, which are presumably separated by a few minutes. The spatial resolution (i.e., 1 pixel) of images (\(200 \times 228\hbox { pixels}\)) is \(\sim 37 \times 31\hbox { km}\) when viewed vertically from the Polar satellite with its apogee of \(9R_{E}\). Practically, UVI images are smeared by approximately \(\pm \,5\) pixels owing to satellite-spin-associated wobbling (e.g., Germany et al. 1998; Frank et al. 2001a). The emission altitude was assumed to be 120 km from the ground. Slant path brightness enhancements were corrected by using an empirical model similar to a cosine curve (e.g., Germany et al. 1998). The emission brightness was converted to the energy flux of the precipitating electrons that cause auroras by using 130 R per \(\mathrm{mW}\ \mathrm{m}^{-2}\), referring to the results of Galand and Lummerzheim (2004).

Figure 2a1 shows an example of the UVI image in the raw charged couple device (CCD) coordinates with an overlaid geographical map. The magnetic coordinates (Fig. 2a2) have often been used to show GIs in previous substorm studies. We calculated the magnetic latitude (MLAT) and longitude (MLON) in degrees and the magnetic local time (MLT) in hours of the modified magnetic apex coordinates (Richmond 1995) for a reference altitude of 110 km using the IGRF-12 (Thebault et al. 2015) model.

All-sky images at Muonio (65 MLAT)

The satellite images were compared with ASIs observed at Muonio (MUO: 64.6 MLAT, 105.2 MLON, \(68.02{^{\circ }}\hbox {N}\), \(23.53{^{\circ }}\hbox {E}\)) in Lapland, Finland (e.g., Fig. 2b1 and b2). The red circle in Fig. 2b2 and a1 with a diameter of \(\sim 1000\hbox { km}\) roughly indicates the field of view of the imager. This intensified CCD all-sky camera is maintained by the Finnish Meteorological Institute (e.g., Syrjäsuo et al. 1998; Sangalli et al. 2011; Partamies et al. 2015).

We used the green-line (557.7 nm) images captured every 20 s with an exposure time of 1 s. Figure 2b1 shows an example image in raw (CCD) coordinates. The \(512 \times 512\) pixel images correspond to \(\sim 1\hbox { km}\) resolution overhead at an assumed emission altitude of 110 km. The geodetic coordinates (Fig. 2b2) have often been used to show ASIs in previous substorm studies.

Satellite-based global images

Figure 3a shows the time sequence of Polar/UVI images in the MLT–MLAT polar coordinates. An auroral brightening was first observed in the panel labeled in red at 2128:07 UT, which is midpoint of the 36.8-s exposure time. The brightening was located around [23.2 MLT, 64.6 MLAT] as indicated by the red circle. The time of the previous image was 2127:30 UT, just prior to the brightening event. We consider the average of these two times, 2127:49 UT, as the beginning of the auroral brightening event.

This brightening appears to be localized at the beginning of the event and was immediately followed by poleward expansion, as shown in later panels. These are typical signatures of auroral breakup, or substorm onset, in GIs (e.g., Liou et al. 2000; Frey et al. 2004). This result is also shown in the auroral keogram (Fig. 3b), where the onset at 2127:49 UT is marked by a solid vertical line. No other brightenings were evident before the onset, as also indicated by the average of the keogram data between 62 and 70 MLAT (Fig. 3c), particularly around 3 min (dashed vertical line) before the onset. Because the possible two-stage development was not identified, it is unclear solely from the GIs whether this substorm onset is the Akasofu substorm onset.

Ground-based all-sky images

In contrast, AIB was observed in the ground ASIs (Figs. 4 and 5) captured at the MUO station. Figure 4 shows the time sequence of full-time 20-s resolution images. Figure 5a shows three selected images that represent moments during the quiet time, initial brightening, and poleward expansion. A brightened auroral arc is evident in the 2126:20 UT image in Figs. 4 and 5a, but it is subjective to determine precisely when it began. Tracing this arc back in time starting at 2126:20 UT, we determined that the arc brightening began at the 2125:00 UT image in Fig. 4. This detailed selection of the start time is moderately supported by the auroral keogram (Fig. 5b) and by the auroral brightness near the onset location (Fig. 5c), which shows a small enhancement in its increase rate.

However, faint spots appear before 2125:00 UT in Fig. 4. In particular, a spot at (23.18 MLT, 64.6 MLAT) near the onset MLT in the 2124:00 UT panel may be another possible candidate for the start of the brightening. Thus, the selection of the 2125:00 UT image as the first brightening is subjective for approximately 1 min. The auroras show faint azimuthally separated structures near the onset MLT, e.g., 2124:40 UT and 2126:20 UT panels. These structures are presumably consistent with auroral beading (e.g., Donovan et al. 2006; Liang et al. 2008), although their signals are weak in this particular event.

The first brightening identified above was centered at [23.2 MLT, 64.6 MLAT] in the 2125:00 UT image and spanned approximately between 22.8 and 23.6 MLT in the 2126:00 UT image. Because this brightening occurred simultaneously within a few minutes across a wide longitude, it can be interpreted to be the AIB that was used to define the substorm onset by Akasofu (1964). It should be noted that we do not specifically require the AIB to be as wide as those illustrated in Akasofu (1964) and Akasofu et al. (2010), which span 4–6 h in MLT and would be typically too wide before the poleward expansion. Because the original images were captured at 20-s intervals, we assumed that the AIB began at 2124:50 UT, 10 s before 2125:00 UT.

The brightened arc shows a small split at 23.1 MLT in the 2127:40 UT image, but the poleward expansion has not yet started in this image and in the keogram. The poleward expansion actually begins in the next image, at 2128:00 UT, when the bright part (22.8–23.5 MLT) of the auroral arc began to split in the northern direction. The resultant poleward arc expanded further poleward, as shown in the 2129:00 UT panel. We assumed that this poleward expansion began at 2127:50 UT, 10-s before 2128:00 UT. An associated auroral brightening occurred simultaneously or in the previous image at 2127:40 UT, depending on the subjectivity. Because this second brightening was followed immediately by the poleward expansion, it is considered in the present study to be an auroral breakup.

In summary, the AIB was identified at 2124:50 UT with a subjectivity of approximately 1 min. Mende et al. (2009) also reported that AIB can be too gradual to identify within \(\sim 10\hbox {-s}\) accuracy. The increasing rate of auroral brightness was approximately constant during the AIB (Fig. 5c). The poleward expansion was identified at 2127:50 UT, which is delayed from the AIB by at least 2 min and most likely 3 min. Thus, the two-stage development was evident in the ASIs.

Comparisons of ground and satellite images

Figure 6 shows simultaneous comparisons of ground and satellite images. Figure 6a shows ground ASIs observed at MUO, projected in the geodetic coordinates. These ASIs were selected with 40–180 s separations to represent the observed instances (a1) before onset, (a2) at the start of the AIB, (a3 and a4) during the AIB, (a5) at the start of the poleward expansion, and (a6) during the poleward expansion.

Figure 6b shows the corresponding Polar UVI images for the same fixed area as that in Fig. 6a. Each image was selected to correspond to an ASI (Fig. 6a) within 7 s. A comparison of Fig. 6a and b reveals the poleward expansion in the ASIs (a5 and a6) was simultaneously observed in the GIs (b5 and b6), although the GIs appear smeared by satellite-spin-associated wobbling. In contrast, the AIB in the ASIs (a2, a3, and a4) was not evident in the corresponding GIs (b2, b3, and b4).

These characteristics were also observed in the keograms (Fig. 7), where slices of images at 23.2 MLT between 62 and 68 MLAT are shown. Again, poleward expansion was observed at about 2127:50 UT both in the (a) ASIs and (b) GIs. In contrast, the AIB (i.e., Akasofu substorm onset), which was observed at 2124:50 UT in the ASIs, was not evident in the GIs until the poleward expansion began.

In summary, the counterpart of the AIB was not evident in the GIs. Consequently, the observed first brightening in the GIs corresponded to the second brightening in the ASIs (i.e., poleward expansion). Therefore, we suggest that the substorm onsets in the GIs and ASIs represent different stages of substorms, particularly when these onsets are identified independently (e.g., Figs. 3 and 4).

Solar wind and geomagnetic indices

Figure 8 shows the solar wind and geomagnetic indices obtained from the Operating Missions as Nodes on the Internet (OMNI) (King and Papitashvili 2005) 1-min resolution data. The north–south component of the interplanetary magnetic field (IMF) was weakly southward between 0 and − 3 nT from 2054 UT, or about 30 min prior to the AIB, to 2145 UT. The dawn–dusk component and the magnitude of IMF were relatively strong during this interval, at 6 nT duskward. The solar wind speed was relatively high at 600 km/s, although the plasma density was relatively low at \(2/\mathrm{cm}^{3}\), resulting in a normal dynamic pressure at 2 nPa.

The geomagnetic condition was moderately disturbed during the 2-h period, as shown by the Kp index (3+ to 3) and SYM-H indices (\(\sim {-}\,30\hbox {nT}\)). This disturbed interval belonged to a co-rotation interaction region-type weak (peak \(\sim {-}\,40\hbox {nT}\)) magnetic storm that began four days prior at around 9 UT on December 3, 1999 (not shown). The AL index began to develop at 2129 UT (Fig. 8), 1 min after the poleward expansion, and 4 min after the Akasofu substorm onset in the ground ASIs (Fig. 7). The AL was − 127, − 128, and − 245 nT at 2127, 2128, and 2129 UT, respectively, and reached its peak value of − 355 nT at 2134 UT.

Negative bays in the ground magnetic field

Substorm onsets are also traditionally identified by using negative bays, positive bays, and Pi2 pulsations in ground magnetic field data. Figure 9 shows the negative bays with the 10-s resolution ground magnetic field data obtained from the International Monitor for Auroral Geomagnetic Effects (IMAGE) project (e.g., Viljanen et al. 1995; Tanskanen 2009). Figure 9a shows the northward (X), eastward (Y), and downward (Z) components of all available data in the geomagnetic coordinates. The Kiruna station (KIR: 64.6 MLAT, 102.7 MLON) was located at 23.4 MLT at the time of the AIB (2124:50 UT, the first red line). This location was close (Fig. 9d) to the AIB centered at [23.2 MLT, 64.6 MLAT]. However, no significant magnetic variations were detected at KIR and at other stations at the time of the AIB.

In contrast, the poleward expansion (Figs. 4 and 5) that began at 2127:50 UT (the second red line) was accompanied by decreases up to \(\sim 400\hbox { nT}\) in the X component. The negative bays began around 2128 UT at the KIR and MUO stations near the onset MLAT, where the bays weakened temporarily after 2129 UT, presumably because the current center had moved poleward. The negative bay was more clearly observed just north (\(65.2{^{\circ }}{-}65.8{^{\circ }}\): ABK and KIL) of the onset MLAT (\(64.6^\circ\)). Stations at higher latitudes (\(66.3{^{\circ }}{-}66.5{^{\circ }}\): AND and TRO) detected sharp negative bays 1 min later at 2129 UT, and detected the largest decrease (\(\sim 400\hbox { nT}\)) among all stations at 2130 UT.

These magnetic field data were used to infer the equivalent electric current at an altitude of 110 km by using the method described in Juusola et al. (2016). We first derived the two-dimensional maps (not shown) of the equivalent current and then focused on the KIR station meridian at \(103{^{\circ }}\) magnetic longitude, which was typically 0.2 h east of the auroral onset MLT center at 23.2 h. Figure 9b shows the time evolution of the inferred equivalent current intensity at this longitude. The equivalent current intensified around the time of the auroral poleward expansion, at 2127:50 UT. This intensification began around the auroral onset MLAT, at \(64.6{^{\circ }}\), and then expanded poleward; these results are consistent with the latitudinal dependences of the observed magnetic variations shown in Fig. 9a.

The major negative bay of \(\sim 400\hbox { nT}\) beginning at 2128 UT in Fig. 9a is considered to be a traditional substorm onset signature in the present study. However, it should be noted that much smaller variations are visible when the vertical scale is changed (Fig. 9c). Decreases in X began at 2124 UT near the onset latitude, \(\sim 15\hbox { nT}\) at KIR and \(\sim 20\hbox { nT}\) at MUO, corresponding to the weak enhancement in the equivalent current intensity at 2124 UT (Fig. 9b). These may be associated with the AIB (2124:50 UT about \(\pm \,1\hbox { min}\)), although it would be too weak to be identified conventionally as a substorm onset.

Positive bays and Pi2 pulsations

Figure 10 shows the 1-s resolution ground magnetic field data obtained through the Sub-Auroral Magnetometer Network (SAMNET) project (e.g., Yeoman et al. 1990), where stations below 60 MLAT were selected. Positive bays in the X component were evident at HAN, NUR, and KVI stations near the onset MLT, at 23.2 h. These positive bays started at about 2128:50 UT, which is \(\sim 1\hbox { min}\) later than the poleward expansion but 4 min later than the Akasofu substorm onset.

Simultaneously, magnetic pulsations began at these stations. The peak-to-peak amplitude was approximately 3 nT with a periodicity of \(\sim 50\hbox { s}\) inside the Pi2 range, at 40–150 s. Thus, the midlatitude Pi2 pulsations were observed in association with the poleward expansion. Associated Pi2 pulsations were observed at other stations (GML, BOR, and YOR), although their beginnings were less clear.

It should be noted that we concentrated on Pi2 pulsations at midlatitudes, where no aurora was observed in the GIs (Fig. 3a). Such Pi2s represent global magnetic variations and thus have been traditionally used as a substorm onset indicator. At auroral latitudes, Pi2-range variations may be observed at the time of the IB if the observatory is coincidentally located at the right place. However, the temporal and spatial variations of such auroral-latitude Pi2s are well correlated with local auroras (e.g., Rae et al. 2009). Thus, the implications of such auroral-latitude Pi2s differ from the lower-latitude Pi2s. Although the global variation component may be included in the Pi2-range variation at auroral latitudes, its extraction is difficult in the presence of auroras.

In summary, the geomagnetic signatures of the substorm onset were observed, including the start of development in AL, negative bay, positive bay, and midlatitude Pi2 pulsation. Such signatures began at about 2128–2129 UT, which is 0–1 min after the poleward expansion at 2127:50 UT but 3–4 min after the AIB at 2124:50 UT. Thus, these signatures do not likely correspond to the AIB; rather, they are more likely to be poleward expansion. The absence of significant geomagnetic responses to the AIB was also reported by Nishimura et al. (2012), Lyons et al. (2013), and Ieda et al. (2016). Lyons et al. (2013) further concluded that significant geomagnetic variations correspond to post-onset streamers from the poleward boundary of the auroral bulge. This detailed correspondence was difficult to confirm in this particular event, with the limited time resolution. Therefore, we conclude simply that significant geomagnetic variations correspond to poleward expansion.

Time delay of substorm onsets between ground and satellite images

In the present study, the substorm onset identified by using GIs was delayed from the ASI data by 3 min. This delay corresponds to the time difference between the AIB and the poleward expansion and thus corresponds to the duration of the first stage (Fig. 1b) of the substorm expansion phase in Akasofu (1964) of a few minutes. This Stage 1 often includes auroral rays (Akasofu 1964). We believe that auroral rays and auroral beads are different views of the same auroral structure and that both can be recognized as detailed features of a longitudinally wide brightening (i.e., AIB in Fig. 11).

The duration of the AIB in the ASIs was 2.5 min (Rae et al. 2009), a few minutes (Mende et al. 2009), and 7 min (Motoba et al. 2014) in previous case studies. The duration was 1–2 min on average and extended to 7 min in a statistical study (Nishimura et al. 2016). Thus, large diversity occurs in the identified delays/durations (\(\sim 1{-}7\hbox { min}\)). In the present discussion, we assumed that the time delay is typically a few minutes.

It is currently difficult to comprehensively understand this diversity, although a clue may be that the AIB tends to have a short duration when it intensifies rapidly (Nishimura et al. 2016). Practically, precursor brightenings are often observed prior to the AIB (e.g., Ieda et al. 2016). It is sometimes difficult to determine whether such a brightening is the AIB or a precursor, particularly when it does not decay significantly, leading to subjectivity in the duration of the AIB. Substorm onsets with a delay/duration shorter than the time resolutions of the GIs would appear to be simultaneous between the ASIs and GIs. Even in such cases, the implications of observed onsets are presumably different between the ASIs and GIs.

The delay of GI onsets from ASI onsets has been assumed to be small, at less than \(\sim 1\hbox { min}\) (e.g., Liou 2010), without direct comparison of GIs and ASIs. Pi2s have been classical substorm onset signatures (e.g., Rostoker et al. 1980; Olson 1999; Nosé et al. 2012). GI onsets have been observed \(\sim 1\hbox { min}\) prior to Pi2s (Liou et al. 2000). This correspondence may verify that the delays of GI onsets from ASI onsets are small. However, the present study and Ieda et al. (2016) suggest that major Pi2s are not likely associated with the Akasofu substorm onset, but rather with the poleward expansion later in the ASIs. Thus, the correspondence of GI onsets (i.e., poleward expansion) to Pi2s does not necessarily imply that the delays of GI onsets from ASI onsets are small. Rather, it suggests that the substorm onsets in the GIs are delayed with respect to the AIB in ASIs by more than that expected, depending on the duration of the AIB.

Causes of differences between ground and satellite images

Poleward expansion was observed in both ASIs and GIs. This sudden change appeared to be even more evident in the GIs (Figs. 6b and 7b) than in the ASIs (Figs. 6a and 7a), indicating that the practical sensitivity of the GIs is sufficient to identify poleward expansion. In contrast, the AIB was not evident in the GIs, indicating that the sensitivity of GIs is considerably less than that of ASIs for identifying the AIB.

These results suggest that the different responses between ASIs and GIs may depend on the latitudinal thickness of the auroras. Our interpretation is that the brightness of the aurora is underemphasized when the target is thinner than the spatial resolution of images. This underemphasis is attributed to the averaging of an area that includes both the thin aurora and the adjacent dark region. The AIB is less evident in GIs, presumably because it is thin in terms of the latitude range, particularly at the beginning, compared with the spatial resolution of GIs. Thus, its brightness would be reduced significantly by area averaging. In contrast, the poleward expansion includes a thickening of the bright aurora; thus, its brightness would be reduced at the beginning but would not be reduced after the expansion has reached the spatial resolution of GIs. That is, the increase in brightness would be overemphasized in GIs when it begins to detect poleward expansion (i.e., auroral breakup).

Another possibility is that these different responses in ASIs and GIs may be attributed to the difference in wavelengths used to observe the auroras. The difference in wavelength did not result in significant differences in the brightness of the poleward expansion. However, it may contribute to difference in the brightness of the AIB. Both satellite (170 nm) and ground (557.7 nm) images are expected to be sensitive to precipitating electrons in the keV range. Thus, the difference in wavelength likely did not contribute significantly to the difference in observed auroras if the onset was dominated by keV-range electrons. However, precipitating electrons may belong to other energy ranges for the AIB. In such cases, the difference in wavelength may contribute to the different responses.

The AIB was not evident in the GIs in the present case; in other cases, the AIB may be sometimes visible in GIs depending on the conditions of auroras and cameras. However, the wide brightening is not explicitly included in identifications of substorm onset in GIs (e.g., Frey et al. 2004; Liou 2010), although it is not explicitly excluded. Thus, the AIB has not been typically recognized in GIs thus far. The AIB would be difficult to recognize as a substorm onset (i.e., sudden brightening) in GIs, not only because its brightness is underemphasized, but also because the increase in brightness of the following poleward expansion is overemphasized. With these assumptions, it may be sometimes possible to recognize a weak brightening in GIs as belonging to the AIB a few minutes prior to the major brightening (i.e., poleward expansion).

Clarifications of substorm onset definitions

We have inferred that the traditionally identified onset brightening in satellite GIs does not necessarily correspond to the Akasofu substorm onset. Instead, it tends to represent the poleward expansion that follows a few minutes later (Fig. 11). Below, we discuss the reason why this interpretation has not been widely recognized.

Impacts on past tail reconnection timing

Reconnection-associated fast plasma flows are often observed in the magnetotail near the time of a substorm onset identified by using Pi2s or GIs (Hones et al. 1984; Moldwin and Hughes 1993; Slavin et al. 2002; Ieda et al. 2008). These fast flows have occasionally been further identified a few minutes prior to the substorm onset (Nagai et al. 1998; Ohtani et al. 1999; Baker et al. 2002; Kepko et al. 2004; Miyashita et al. 2009).

However, such conclusions depend on the definition of substorm onset. Whether the identified substorm onset corresponds to the AIB or poleward expansion in ASIs has not been specified in these previous studies. In the present study, the onsets in Pi2s and GIs corresponded to poleward expansion rather than the Akasofu substorm onset. This result suggests that unless the longitudinally wide AIB was explicitly considered, the substorm onsets identified in past studies did not correspond to the Akasofu substorm onset but rather to poleward expansion.

Fast flows have always been initiated within a few minutes of the isolated auroral breakup in GIs (i.e., poleward expansion) if the satellite was located near the onset MLT (Ieda et al. 2008). However, unobserved AIB may have occurred prior to the auroral breakup (i.e., poleward expansion). Thus, these fast flows may be delayed from the possible AIB, as was reported in a case study by Ieda et al. (2016). In summary, no evidence exists for reconnection-associated fast flows prior to the Akasofu substorm onset. Therefore, the developed reconnection does not likely trigger the Akasofu substorm onset.

Reconnection-associated fast flows may be associated with auroral streamers. Some brightenings (e.g., 2126:17 UT panel) occurred near 73 MLAT near the onset MLT sector in Fig. 3a and b. Interestingly, an auroral streamer was formed at 72 MLAT near the onset MLT simultaneously with the breakup (2128:07 UT panel). This simultaneous occurrence may be a coincidence, or it may suggest that the auroral breakup (i.e., poleward expansion) and tail reconnection occur simultaneously.