On the Unexpected Correlation Between Relativistic Electron Microbursts and Patchy Pulsating Aurora

In this letter, we present the results of a conjunction between the SAMPEX satellite and a THEMIS all-sky imager in Gillam, Canada—showing a high correlation between relativistic, > 1 MeV, electron microbursts and a type of pulsating aurora called patchy aurora. The correlation was 0.8, and is not serendipitous. While the relationship between pulsating aurora and 10-100s keV microbursts has been previously predicted, here we show a strong association between keV and MeV electron dynamics— possibly spanning two orders of magnitude. Importantly, this result shows that the dynamics of relativistic radiation belt electrons are at times intimately tied to keV electron precipitation, and can not be studied in isolation.

two orders of magnitude. Importantly, this result shows that the dynamics of relativis-23 tic radiation belt electrons are at times intimately tied to keV electron precipitation, and 24 can not be studied in isolation. 25 Plain Language Summary 26 In this letter we present a coordinated observation between a low Earth orbiting satel-27 lite, orbiting at 400 km altitude above Earth's surface, and an auroral all-sky imager in 28 Canada. This observation showed a connection of a type of pulsating aurora, called patchy 29 aurora, with extremely energetic and intense bursts of electron radiation called microbursts. 30 This link is surprising because the electron energies responsible for auroral light are 100 31 times lower than the electrons that were directly observed in space. Our result implies 32 that the mechanism responsible for patchy aurora and microbursts is likely the same, 33 and could be capable of affecting electrons with vastly different energies. This result is 34 a major step towards unifying the microburst and patchy aurora phenomena and shows 35 that the dynamics of high-energy electrons located in near-Earth space can be intimately 36 tied to much lower energy electron precipitation, and must therefore be studied together. Energetic electrons in Earth's magnetosphere are highly dynamic, governed by a 39 complex interplay of many source and loss mechanisms (e.g. Ripoll et al., 2020, and ref-40 erences therein). One important loss mechanism is the interaction between electrons and 41 plasma waves, resulting in precipitation of electrons into Earth's upper atmosphere (e.g into pulsating aurora in 1979 directly observed pulsating aurora electrons in the > 140 66 keV integral energy channel (Sandahl et al., 1980). More recently, Miyoshi et al. (2015) 67 estimated the upper energy bound of pulsating aurora using an ASI and a very high fre-68 quency radar at Tromsø, Norway; pulsating aurora was observed by the ASI, while the 69 radar observed a substantial electron density enhancement at altitudes as low as 68 km-70 corresponding to 200 keV electrons (e.g. Fang et al., 2010). These authors concluded that 71 the keV pulsating aurora can be associated with sub-relativistic and relativistic electron 72 precipitation. Together, these observations suggest that the electrons that produce pul-

94
Prior studies have also reported another form of electron precipitation in this en-95 ergy range (10s KeV-MeV) called electron microbursts. Microbursts are typically defined 96 as a subsecond intense increase of energetic electron precipitation into Earth's atmosphere 97 and are also believed to be scattered by chorus waves (Breneman et al., 2017;Lorentzen, 98 ships, pulsating aurora and microbursts, especially relativistic microbursts, are often stud-116 ied independently, and only occasionally their relationship is considered. An exception 117 is Hofmann and Greene (1972). The authors used data from a balloon that carried a pho-118 tometer and a scintillator that showed correlation between violet light and > 20 keV 119 X-ray modulation on both the on-off and modulation time scales. The upper energy of 120 the precipitation was 50 keV as the > 50 keV scintillator channel did not respond. Hofmann 121 and Greene (1972) classified the subsecond > 20 keV X-ray modulations as microbursts.

122
The link between pulsating aurora and relativistic microbursts is observationally 123 elusive, but has been theoretically proposed. Recent test-particle simulations by Chen is unreported. This is likely due to several factors including difficulties finding good con-129 junctions between satellites and ground-based imagers capable of observing both phe-130 nomena, and the small size of individual pulsating aurora patches and microbursts (Johnstone, 131 1978;Blake et al., 1996;Shumko et al., 2020;Grono & Donovan, 2020).

132
In this letter we present observational evidence supporting a link between pulsat-133 ing aurora and relativistic (> 1 MeV) microburst electron precipitation. We showcase

THEMIS ASI
The 256×256 pixel images are taken at a 3-second cadence: a 1-second exposure 163 followed by 2-seconds to process the image. To accurately compare the SAMPEX and 164 THEMIS ASI time stamps, we highlight that the ASI times are recorded at the start of 165 each exposure. To analyze the white light images, we used the THEMIS ASI skymap cal-166 ibration data provided by the University of Calgary. The calibration data contain, among 167 other things, arrays that map each pixel to an elevation and azimuth.

168
In this case study, we used the THEMIS ASI camera stationed in Gillam, Canada,   In essence, our analysis involves locating SAMPEX's footprint in each auroral im-182 age, estimating auroral intensity at this footprint, and comparing this auroral intensity 183 with the > 1 MeV electron counts observed by SAMPEX. We did this with the follow-184 ing four steps.

185
Step one: we interpolated the 6-second SAMPEX (latitude, longitude, altitude) co-186 ordinates to the 3-second Gillam time stamps. This is important because when SAM-187 PEX was at high elevations in the Gillam images, it rapidly traversed the field of view, 188 thus resulting in large uncertainties in the estimated auroral intensity at the footprint.

189
Step two: we calculated SAMPEX footprint's location in each image. The  PEX position at ≈ 400 km altitude was magnetically mapped to the approximate pul-

197
Step three: we detrended the HILT count time series to reduce the effects of the 198 SAMPEX spin. This was necessary in order to correlate only the precipitating > 1 MeV 199 microburst electrons with the aurora. We isolated the precipitating microburst electrons

203
Step four: for each image we calculated the mean auroral intensity in a 10 × 10 204 km box around the SAMPEX footprint. We then compared the auroral intensity to the  Fig. 1b show that the PA persisted for minutes at a time 220 while it changed shape and drifted to the East. The patchy aurora does not appear to 221 be pulsating, but that could be due to instrumental effect that will be discussed later.

222
Consequently, patchy aurora is classified as a type of pulsating aurora but is bewilder-223 ingly distinct from patchy pulsating aurora. Grono and Donovan (2020) argue that patchy 224 aurora and patchy pulsating aurora are "closely related in terms of the underlying scat-225 tering mechanism responsible for the precipitation."  HILT counts and the ASI intensity is 0.8.

249
To check that this correlation is a unique signature and not a coincidence, we con-250 duct a test using the above methodology with one change: we offset (i.e. lagged) the ASI 251 frames in time. Put simply, this test quantifies how well the microbursts observed by HILT 252 correlate with the aurora observed at a different time, but along the same footprint path. 253 Figure 3 shows the test result for ASI frame lags between ±10 minutes. The HILT time 254 series correlated well only for the PA that drifted through the SAMPEX footprint for 255 ≈ 5 minutes (correlation was > 0.6 for 2 minutes during that time interval).

256
For the final check, we investigated the effect of different mapping altitudes. We 257 repeated the above analysis with varying auroral emission altitudes spanning 80-120 km. 258 We found that 100 km altitude maximized the correlation at 0.81, and correlation re-  titudes of 80-120 km, and the correlation is significant for only that PA and no other au-266 rora observed within 10 minutes. Thus, this correlation is unlikely due to chance.

267
Now we infer the lower energy bound of the precipitating electrons using the MSP 268 data. Panels c and d in Fig. 1 show that the green and blue channel luminosity-but 269 not the red channel luminosity-was elevated near zenith at 11 UT. In principle, the three  Kataoka et al., 2012;Nishiyama et al., 2012). Nevertheless, we can confidently say that 304 some of the aurora toggled on and off between sequential frames in Movies S1 and S2, 305 so at least some of the aurora was aliased to a 6-second period. 306 We stress that the correlation between pulsating aurora and MeV electrons does 307 not imply that the visible pulsating aurora light is emitted by MeV electrons. But this 308 result does imply that in the pulsating aurora phenomenon, aurora-producing keV elec-

314
Now we speculate on the scattering mechanism. Our biggest assumption is that  Movie S1 A 20-minute fisheye view from the Gillam THEMIS ASI camera. North is towards the top and east is to the right. The camera observed a combination of amorphous and patchy pulsating aurora.
Movie S2 Conjunction movie. Top panel: a 150-second long movie that shows the same fisheye lens view as Movie S1, but with the SAMPEX 90 km footprint superposed. The large red dot was SAMPEX's instantaneous footprint, and the yellow box outlines the 10 × 10 km area used to calculate the mean ASI intensity in Fig. 2 in the main text.
Middle panel: the SAMPEX-HILT time series, showing > 1 MeV precipitation. The sinusoidal count oscillation was due to SAMPEX spinning (and was detrended in the bottom panel). Two distinct and rapid sequences of microbursts (also known as microburst trains) were observed: one at approximately 11:00:00 UT, and the other at 11:00:14 UT.
Interpretation of the MSP ratio data The intensity ratios among different auroral emission lines have long been used in inferring the characteristic energy of auroral precipitation (e.g. Rees & Luckey, 1974;Meier et al., 1989;Grubbs et al., 2018;Liang et al., 2018;Hecht et al., 2006). The green-to-blue ratio is most used in the case of energetic electron precipitation. In short, the lower the green-to-blue ratio, the higher the characteristic energy. Note that while the 470.9 nm emission is actually measured in realistic MSP data, we convert it to the 427.8 nm emission intensity in our following presentation to facilitate comparison with published results of green-to-blue ratios, since most existing auroral models calculate the 427.8 nm emission. The 427.8 nm and 470.9 nm emissions both belong to the N 2+ 1NG system; a I 427.8 /I 470.9 ratio of ≈ 4.97 (which we use in : X -3 our analysis) exists between them under typical ionospheric conditions (e.g Jones, 1975;Shepherd et al., 1996). here approximately represent the green-to-blue ratios of the patchy pulsating auroras with the ambient diffuse auroral components subtracted. These ratios, based upon evaluations from existing auroral models (e.g. Rees & Luckey, 1974;Grubbs et al., 2018), and within the reasonable range of the atmospheric N2/O content ratio (Hecht et al., 2006), would point to that the characteristic energy of the precipitation responsible for the patchy pulsating auroras is very likely > 10 keV.