A preview of this full-text is provided by Wiley.
Content available from Journal of Geophysical Research: Space Physics
This content is subject to copyright. Terms and conditions apply.
1. Introduction
A nightsky optical phenomenon named Strong Thermal Emission Velocity Enhancement (STEVE) has
been a topic of great research interest in the past 1–2years. Its presence was reported in the literature over
a century ago (see a historical review by Hunnekuhl and MacDonald[2020]), but escaped the attention
of scientists for a long time, until it was brought forth again by citizen scientists and auroral photography
enthusiasts in recent years. To date, while its generation mechanism remains unknown, some important
advancements have been made toward the understanding of STEVE. (1) STEVE is located equatorward of
traditional auroras, and there has been no evidence of appreciable electron/proton precipitation that could
be the cause of the luminosity (Gallardo-Lacourt etal.,2018a, 2018b; MacDonald etal.,2018), suggesting
that STEVE is not traditional aurora. (2) STEVE is found to be co-located with strong ionospheric elec-
tron heating and fast subauroral ion drifts (SAID) in joint optical and in situ satellite observations (Archer
etal.,2019a; Chu etal.,2019; MacDonald et al., 2018; Nishimura et al., 2019; 2020a). (3) STEVE occurs
during substorm intervals, and typically starts to emerge ∼1h after the substorm onset (Gallardo-Lacourt
etal.,2018a). (4) The emission altitude of STEVE is typically at ∼210–260km, though sometimes a separate
yet weaker STEVE arc may co-exist at lower altitude (Archer etal.,2019b; Hunnekuhl and MacDonald 2019;
Liang etal.,2019). (5) Recently, using spectrograph data it was unveiled that STEVE's main source of bright-
ness comes from an overall enhancement of a continuous visual spectrum, that is, an airglow continuum
(AGC, Gillies etal.,2019; Liang etal.,2019), with small or none out-of-background green-line 557.7nm
emissions contained in the STEVE spectrum. While the existence of night AGC has been known for dec-
ades (Barbier etal.,1951), their commonly reported intensities in the existing literature (e.g. Sternberg &
Abstract Strong Thermal Emission Velocity Enhancement (STEVE) is a nightsky optical
phenomenon of great research interest in recent years. Recent findings indicated that STEVE likely
represents certain extremely intensified chemiluminescence airglow instead of traditional aurora. In
this study, we investigate the patterns and variations of the neutral wind and temperature before the
STEVE emergence using joint scanning Doppler imager (SDI 630nm) and optical all-sky imager (ASI)
observations, and make an initial effort to explore the potential preconditioning role of neutral winds in
the STEVE production. Neutral winds enhance in westward and southward directions following substorm
auroral intensification, and show an equatorward propagating trend from auroral latitudes. However,
in STEVE events the enhanced equatorward winds feature a steep stop/reversal at certain subauroral
latitude, and strong wind convergence is developed there. This pattern sustains for ∼15–20min, and then
STEVE arises at about this stop latitude. The strength of the southward wind intensification and wind
convergence is in general weaker or absent in nonSTEVE substorm events. We propose that enhanced
equatorward winds may transport relevant neutrals species that are key to the STEVE airglow production
to subauroral latitudes, and pile up at the stop latitude of the equatorward winds due to the strong
convergence there. Such a transport/pileup effect led by the neutral winds may prepare a reservoir of
neutral constituent which, when further aided by subauroral ion drift, leads to a dramatic increase of the
airglow production and the STEVE occurrence.
LIANG ET AL.
© 2021. American Geophysical Union.
All Rights Reserved.
Neutral Wind Dynamics Preceding the STEVE
Occurrence and Their Possible Preconditioning Role in
STEVE Formation
Jun Liang1 , Y. Zou2, Y. Nishimura3 , E. Donovan1, E. Spanswick1 , and M. Conde4
1Department of Physics and Astronomy, University of Calgary, Calgary, Alberta, Canada, 2Department of Space Science,
University of Alabama in Huntsville, Huntsville, AL, USA, 3Department of Electrical and Computer Engineering
and Center for Space Sciences, Boston University, Boston, MA, USA, 4Department of Physics, University of Alaska
Fairbanks, Fairbanks, AK, USA
Key Points:
• We investigate the variations of
the neutral wind and temperature
before the Strong Thermal Emission
Velocity Enhancement (STEVE)
emergence using joint scanning
Doppler imager and optical all-sky
imager observations
• Winds enhance in westward/
southward directions following
substorm, but the southward winds
feature a stop latitude before STEVE
emergence
• The neutral wind may lead to
the transport/pileup of neutral
constituents at subauroral latitudes
and contribute to STEVE airglow
production
Supporting Information:
• Supporting Information S1
• Movie S1
• Movie S2
• Movie S3
Correspondence to:
J. Liang,
liangj@ucalgary.ca
Citation:
Liang, J., Zou, Y., Nishimura, Y.,
Donovan, E., Spanswick, E., & Conde,
M. (2021). Neutral wind dynamics
preceding the STEVE occurrence
and their possible preconditioning
role in STEVE formation. Journal of
Geophysical Research: Space Physics,
126, e2020JA028505. https://doi.
org/10.1029/2020JA028505
Received 21 JUL 2020
Accepted 19 DEC 2020
10.1029/2020JA028505
RESEARCH ARTICLE
1 of 30
Journal of Geophysical Research: Space Physics
Ingham,1972; Gadsden & Marovich,1973) are much weaker than that of STEVE (Gillies etal.,2019; Liang
etal.,2019)—STEVE is visible even to naked eyes.
It is fair to state that the above existing findings strongly suggest that STEVE likely stems from certain ex-
tremely intensified chemiluminescence in the upper thermosphere, instead of traditional aurora. In chemi-
luminescence, its brightness is proportional to the densities of the air constituents involved in the chain of
chemical reactions leading to the chemiluminescence (e.g., Young,1969). The fact that STEVE occurs at
altitudes higher than 200km excludes the possibility that STEVE be purely neutral airglow without much
interaction with plasma, otherwise it would tend to occur in much lower atmosphere where the neutrals are
much denser. On the other hand, Since the STEVE arc is found as located in a low electron density “trough”
region of the subauroral ionosphere (Archer etal.,2019a; Chu et al., 2019; MacDonald et al., 2018), the
out-of-background enhancement of the STEVE chemiluminescence is not likely to be attributed to a strong
local plasma density, but is more likely to be related to enhancements of neutrals there, at least for some
key constituents relevant to the chemiluminescence. While the underlying mechanism of STEVE remains
unclear to date, and it is not the aim of this paper to investigate in detail the specific STEVE mechanism, the
above observations and considerations point to a gross perspective that, STEVE likely stems from a drastic
plasma-neutral interaction. The ionospheric plasma with reduced density but with high electron tempera-
ture and convective flow speed (i.e., SAID), together with enhanced densities of relevant neutral constitu-
ents, conspire to make the STEVE occur. Partly owing to the above thoughts, the NO2 continuum (Gadsden
& Marovich,1973; Hedin et al., 2012; Sternberg & Ingham, 1972) via a radiative recombination reaction
2
NO O NO hν
has been deemed one of the leading candidates of the STEVE AGC to date by some
researchers including ourselves. While the NO2 continuum itself is a neutral process, the NO production
in the thermosphere is well known to be sensitive to the magnetosphere-ionosphere-thermosphere interac-
tions in numerous ways (e.g., Barth etal.,2003, 2009; Campbell etal.,2006; Lin etal.,2018; Lu etal.,2010;
Solomon etal.,1999; Zhang etal.,2014; Zipf etal.,1970). An enlightening work in this regard was recently
done by Harding etal.(2020), who proposed that the nitrogen molecules may be excited to higher vibra-
tional levels via the collision with fast-drifting ions in SAID. These excited N2 may overcome the activation
barrier and react with oxygen atoms to produce NO, and in turn generate the NO2 continuum that accounts
for the STEVE AGC. However, their proposal still summons a substantial enhancement of local neutral den-
sity as a prerequisite for the mechanism to produce enough AGC yield to account for the STEVE brightness.
While direct in situ neutral observations at STEVE altitudes are not available at the moment, ground-based
measurements of neutral winds and temperature may still offer useful clues to the neutral conditions that
potentially contribute to STEVE. This becomes the motivation and the main goal of the current study. To the
authors' knowledge this is the first study of the neutral atmosphere in which STEVE develops. We shall first
make a key clarification here. As afore-mentioned, based upon existing knowledge there is little doubt that
STEVE embodies complex and interactive magnetospheric-ionospheric-thermospheric coupling processes.
This study is not intended for an investigation of the thermospheric effects led by STEVE. Instead, we shall
focus on the neutral wind/temperature patterns and variations before the STEVE emergence, based upon a
premise that those precedent variations might precondition the thermosphere and lead to a change of rele-
vant neutral constituents, for the ensuing rise of STEVE. It is important to note that such a preconditioning
effect, if indeed contributing to STEVE appearance, cannot be instantaneous considering the usually long
variation timescale of neutrals in the upper thermosphere (e.g., Nishimura etal., 2020b). Neutral wind
variations would presumably occur in advance and persist for a while to let relevant neutrals transport and
buildup, so that the corresponding chemiluminescence airglow may gradually grow to detectable bright-
ness, that is, exceeding certain visual/instrumental threshold. In a statistical study, Zou etal.(2020) found
that the substorm-associated neutral wind variations often commence right after the onset and reach a
quasisteady state within one hour of the onset (see also Cai etal.[2019]). This was also found as compatible
with the calculated ion-neutral coupling timescale (Nishimura etal., 2020b). Coincidently, Gallardo-La-
court etal.(2018b) showed that STEVE typically emerges ∼50–80min after the substorm onset. This time
sequence is consistent with our above hunch.
This paper is organized as follows. Section2 will introduce the instruments used in this study. In Section3,
we shall present two STEVE event examples, with a focus on the characteristic pattern of neutral winds in
tens of minutes before the rise of STEVE, and compare with nonSTEVE substorm events. In Section4 we
LIANG ET AL.
10.1029/2020JA028505
2 of 30
Journal of Geophysical Research: Space Physics
shall discuss the potential implication of our results to the possible mech-
anisms of STEVE. Section5 concludes the paper.
2. Instruments
The neutral measurements in this study come from the scanning Doppler
imager (SDI). A detailed description of the basic principles and opera-
tions of SDIs was discussed by Conde and Smith (1995, 1998) and more
recently by Anderson etal.(2012a) and Dhadly etal.(2015). In a nutshell,
SDI is an all-sky imaging, wavelength-scanning Fabry-Perot spectrome-
ter. It records the spectra of 630 and 557.7 nm optical emissions at over
a hundred locations across the sky, from which the line-of-sight (LOS)
component of neutral winds and the neutral temperature can be derived
from the Doppler shift and Doppler broadening of the emission spectra,
respectively. In line with the fact that typical emission altitudes of STEVE
are at >200km, only 630nm SDI data are used in this study, and we as-
sume a 250km altitude in mapping the data. The SDIs at Poker Flat Range
(PKR, 65.12°N, 147.43°W Geo.) and HAARP (HRP, 62.39°N, 145.15°W
Geo.) are included in this study. In particular, the HRP SDI is roughly col-
located with the Gakona THEMIS ASI, constituting a desirable conjunc-
tion geometry for our research interest. Figure1a shows the geometry of
the field-of-view (FoV) of Gakona ASI, overplotted with the data points
of PKR and HRP SDIs, in an altitude-adjusted corrected geomagnetic
(AACGM) coordinate (Baker & Wing,1989). The SDI LOS velocities are
inverted to horizontal velocity vector maps via a monostatic fit technique
which assumes that vertical winds are constant across the FOV and, for
the software version used in this study, that the zonal gradient of me-
ridional winds is negligible (Anderson etal.,2012a; Dhadly etal.,2015).
Note that such an assumption is not required when LOS data from two
or more stations are involved, as we shall utilize to calculate the wind
divergence in Section3.1. The monostatic fit technique is of course not
without uncertainties. Known inaccuracy would occur at near-overhead
data points where the viewing geometry is bad for estimating horizontal
winds, and near the low-elevation edge where the data points are sparsely
distributed and the oblique LOS integral effect may bring uncertainties.
That said, on overall the fitting technique is usually found to yield a close
approximation to the actual wind field (Anderson etal., 2012a; Dhadly
etal.,2015). The SDI monostatic data and summary plot can be found in
http://sdi_server.gi.alaska.edu/sdiweb/. In the course of our event selec-
tion, we have carefully checked the summary plots and reject events with
suspicious wind vector maps. Cloudy nights are discerned visually based
on collocated all-sky-camera optical observations as well as keograms of
the SDI emission intensity: if the intensity appears overall blurry across
latitudes, the emission has probably been scattered by clouds, and the
event is dismissed. In particular, a good viewing condition over the Ga-
kona all-sky imager (ASI) is a necessary condition of this study. Also, in
this study the zenith and the innermost annulus data points, as well as the data points with elevation angle
<20°, are all excluded for any quantitative use. In the STEVE event with both PKR and HRP SDI data, we
also apply the bistatic technique (Anderson etal.,2012a) to derive the horizontal winds to aid this study.
The THEMIS whitelight ASI (Donovan etal.,2006) will be used in this study for several purposes. Most
importantly, the ASI at Gakona station (GAKO, 62.40°N, 145.16°W Geo.) is used to identify STEVE events
or nonSTEVE events. A 250km emission altitude is assumed for STEVE, which is the same as that ap-
plied for the 630nm SDI data. Liang etal.(2019) reported the possible occurrence of double-layer STEVE
LIANG ET AL.
10.1029/2020JA028505
3 of 30
Figure 1. (a) The observation geometry of the FoV of Gakona all-sky
imager (ASI), overplotted with the Poker Flat Range (PKR) SDI data points
(red) and HRP (green) 630nm scanning Doppler imager (SDI) data points,
in altitude-adjusted corrected geomagnetic (AACGM) coordinate for the
year 2010. Two triangles mark the PKR (higher latitude) and HRP (lower
latitude) station center. The grayed points denote the zenith and innermost
annulus data of each SDI that are ignored in our quantitative analysis of
the wind. (b) The THEMIS ASIs and magnetometer stations involved in
this study. GAKO ASI is highlighted. The PKR site where the PKR SDI and
Poker Flat Incoherent Scatter Radar (PFISR) are located are also labeled.
For station name legends, see Donovan etal.(2006). In both (a and b) the
map coordinate is AACGM.
Journal of Geophysical Research: Space Physics
structures in some events, and evaluated that the lower-altitude STEVE likely centers around ∼150 km
height. However, such a lower-altitude STEVE, even if visibly co-existing, is usually weaker than the high-
er-altitude STEVE which is typically within ∼220–270km altitude (Liang etal.,2019). Therefore, if only
one STEVE arc is seen in the ASI FoV, it is reasonable to assume that it represents the high-altitude STEVE.
Note that in a supplementary material (FigureS2f) we perform a triangulation analysis and evaluate the
emission altitude of STEVE in the March 26, 2008 event as being around 250km. In the two STEVE events
we are to report, the STEVE arc, at least during its initial emergence in the SDI FoV, are fairly close to the
ASI center, so that the error led by the emission latitude uncertainty is presumably small compared to the
latitudinal resolution of the SDI measurement. THEMIS ASIs over Alaska and the western Canada sector,
together with the co-located THEMIS magnetometer stations, are also involved in this study, mainly serving
to infer the substorm context of investigated events, such as the substorm onset time. Figure1b shows the
THEMIS ASIs and magnetometer stations involved in this study. For station name abbreviations, see Dono-
van etal.(2006). The digital meridian spectrometer at PKR and all-sky cameras at GAKO, both operated by
the University of Alaska, Fairbanks are also browsed in some events.
Besides, The Poker Flat Incoherent Scatter Radar (PFISR, 65.1°N, 147.5°W Geo.) data are used in two STE-
VE events to characterize the plasma flow intensification observed in connection with auroral activity after
substorm onset near the equatorward boundary of auroras. PFISR is nearly collocated with PKR SDI. In
April 4, 2010 event, PFISR was run in the Aurora & Convection experiment mode, and its long pulse mode
data with 1-min integration time are used in this study. In March 26, 2008 event, PFISR was run in the IPY04
experiment mode, and its long pulse mode data with 5-min integration time are used. The plasma convec-
tion vector is derived from the long-pulse LOS velocity measurements by assuming that the velocity vectors
are homogeneous in the east-west direction across the radar FoV (Heinselman & Nicolls,2008).
3. Observations
The two STEVE events are from the list in Gallardo-Lacourt etal. (2018b). Readers are referred to their
paper for detailed criteria in identifying STEVE from ASI observations. Some of the key characteristics
include: the STEVE must be equatorward of, and detached from, the main aurora zone; the STEVE arc
is narrow in latitudinal width (∼0.1°-0.2°); the arc and/or the fine structures along it show fast westward
propagation (a few km/s). In all the following figures with ASI images and/or SDI winds, the AACGM
coordinate is used to define the magnetic latitude (MLAT) and longitude (MLON), and to present the data.
3.1. April 4, 2010 Event
This event occurred during a minor geomagnetic storm interval (SYM-H∼−10 to −20) but the substorm
intensity is strong (AL index reaching∼∼810nT). From available optical data, the onset started at the Fort
Simpson (FSIM, 67.3° MLAT, −65.0° MLON) sector at ∼06:18UT. The top panel of Figure2 shows the
keogram of GAKO THEMIS ASI sampled along its center magnetic meridian (−90° MLON). A default emis-
sion altitude of 110km is assumed here which applies to auroras (Whiter etal.,2013)—but not to STEVE,
which we shall use a different emission altitude later. The event occurred in early evening hours. The GAKO
THEMIS ASI was turned on at 06:16UT, yet there was still some apparent twilight contamination in the
western portion of the ASI FoV by that time. Nevertheless, the intensification after 06:20UT and the equa-
torward expansion of auroras after 06:35UT can still be seen. The STEVE trace of our interest is labeled,
which is at distinctly lower latitudes than auroras. The STEVE is a bit faint in this plot since the color scale
here is for substorm auroras; it will become more easily observable in the following plots. The rest of the
panels in Figure2 display a series of images from GAKO THEMIS ASI, showing the STEVE evolution. Note
that a 250km emission altitude is assumed here since we are to focus on STEVE. Starting from ∼07:20UT,
a STEVE arc emerges from the eastern portion of the ASI FoV; it then expands westward and enters the FoV
of HRP and PKR SDIs (station centers marked by triangles) by ∼07:24UT, with an overall mean westward
expansion speed of ∼3km/s. The potential process underlying the fast westward propagation of STEVE will
be discussed in Section4. The STEVE arc then slowly migrates equatorward. Based on the FoVs of available
THEMIS ASIs, this STEVE event is relatively short-lived and vanishes at ∼07:40UT. Readers are referred
to a supplementary movie to view the full STEVE evolution. In our following presentation, when we refer
LIANG ET AL.
10.1029/2020JA028505
4 of 30
Journal of Geophysical Research: Space Physics
to the “emergence time” of STEVE, we mean the time when STEVE locally arises in the FoV of SDIs, since
that is where our available neutral measurements are confined.
In Figure3, we plot side-by-side the observations from PKR (leftside) and HRP (rightside). Their top panels
show the GAKO THEMIS ASI keogram sampled along −93.6° (−90°) MLON, which is the center longitude
of PKR (HRP) SDI, assuming a 250km emission altitude. Since the STEVE is fairly close to the image center
in this event, the error led by the emission height uncertainty is trivial. In a supplementary material (Fig-
ureS1) we also show the keograms based upon 220 and 270km emission altitudes, and note that the STEVE
displacement led by different assumed emission heights is very small and will not affect our analysis and
results, considering the latitudinal resolution of SDI. The second panel of each subfigure, which shows the
east-west convection velocity deduced from PFISR observations, is duplicate. The rest of the panels show
LIANG ET AL.
10.1029/2020JA028505
5 of 30
Figure 2. The top panel shows the keogram of GAKO ASI based on 110km emission height. The rest of panels display
a series of GAKO ASI images with emission height of 250km, showing the evolution and westward propagation of
Strong Thermal Emission Velocity Enhancement (STEVE) for April 4, 2010 event. The locations of HRP (lower-latitude
one) and PKR (higher-latitude one) SDIs are marked as triangles. Images are shown in AACGM coordinates.
Journal of Geophysical Research: Space Physics
the keograms made from the PKR and HRP SDI data. In making these keograms, we adopt the monostatic
fit data from the two SDIS and sample the relevant parameters (temperature, velocity) at each latitude bin
within ±3° MLON around the PKR (HRP) center meridian. The station latitudes of PKR and HRP SDIs
are ∼65.8° and 63.5° MLAT, respectively. Note that for all the SDI keograms we have also overplotted the
STEVE trace obtained from the ASI keogram sampled at the same PKR/HRP meridian (top panel) to fa-
cilitate a comparison with the STEVE emergence time and latitude. We have also checked the PFISR flow
and HRP SDI wind data at earlier times (not shown) to obtain a broader context of the convection and
wind evolution. There was an earlier and smaller substorm onset at ∼04:55UT inferred from available op-
tical and magnetometer data, and westward plasma flows of several hundred m/s had been seen on PFISR
since ∼05:00 UT. However, noticeable westward and southward wind enhancements occur only after the
much stronger substorm with onset at ∼06:18UT. We of course do not deny that, due to the relatively long
response timescale of neutral winds to ion drag, the preexisting westward plasma flows might also partly
contribute to the later wind development. Following this substorm intensification and expansion, strongly
enhanced westward flows (>1km/s) appear at ∼65°.5-67° MLAT. Note that ∼65.5° MLAT is the equator-
ward edge of the PFISR FoV, so that there is unfortunately no plasma flow measurement at lower latitudes.
The third panel of Figure3 shows the neutral temperature (Tn) measurements from PKR and HRP SDIs.
Note that Tn is plotted in different color scales for two SDIs. We shall first remind that the interpretation of
LIANG ET AL.
10.1029/2020JA028505
6 of 30
Figure 3. The left-side is for PKR and the right-side is for HRP 630nm SDI observations. Their figure formats are the same. The panels from top to bottom
are indexed to facilitate reference. The top panel shows the ASI keogram sampled along the center meridian of PKR/HRP based on 110km emission height.
The second panel shows the east-west convection velocities (positive westward) from PFISR measurements. The third panel shows the neutral temperature
measurements from PKR/HRP SDI. The bottom two panels of each subfigure show the east-west (positive westward) and north-south (positive southward)
components of the neutral wind velocity observed by PKR/HRP SDI. From the third to the bottom panel, a black dotted curve marks the trace of the STEVE arc.
Journal of Geophysical Research: Space Physics
the Tn variations seen on SDI measurements is often complicated by the uncertainty of emission heights. In-
ferred from the NRLMSISE-00 empirical model, Tn may increase by 40–50K over 200–250km altitude range
where the 630nm auroras/airglows may peak. The PKR SDI indicates an equatorward propagation of Tn en-
hancement that originates from ∼67° MLAT. Since such an enhancement is apparently associated with the
PFISR westward flow intensification, and that an equatorward propagation of westward wind enhancement
is jointly seen (to be detailed subsequently), we conceive such a Tn enhancement and its equatorward-prop-
agating trend to be real, instead of being an alias of varying 630nm emission heights. On the other hand, Tn
enhancements at latitudes lower than 63° MLAT are seen on HRP SDI observations. The increment magni-
tude of such enhancements is between ∼60–90K, so that they are unlikely to be fully attributed to the varia-
tions of 630nm emission heights. One may notice the significantly different pattern of Tn seen on two SDIs.
This is of course partly owing to the different latitudinal coverage and longitudinal sampling region (in
making the keogram) of the two SDIs. However, a close examination of the data indicates that, even in their
common area of observations, Tn seen by the two SDIs still show different patterns and absolute values. We
point out that such a discrepancy is not entirely unexpected. Anderson etal.(2012a) compared the Tn meas-
urements from PKR and HRP SDIs in their common volumes of observations, and noticed a correlation as
low as 0.53 and absolute difference as large as ∼100K (see their Figure6). Part of the Tn discrepancies may
be attributed to the intrinsic uncertainty of SDI temperature measurements (up to ∼4%). Another likely
reason for the discrepancy is that the latitudinal-altitudinal distribution of Tn intensifications may cause
different integral effects at different viewing aspects. The lower-latitude Tn intensification shown on HRP
SDI is not well seen on PKR in this event, likely because that structure is near the lower-elevation edge of
the PKR SDI where a latitudinal smearing effect along the LOS direction is expected. For the higher-latitude
Tn intensification structure seen on PKR SDI, we speculate that the altitude ranges of the Tn enhancement
there (presumably led by Joule heating) might be below the peak height of the auroral 630nm emissions.
In an oblique view from the HRP SDI, the contribution from this Joule heating region is overridden by the
contributions from the main auroral 630nm emission region at higher latitudes and altitudes, so that the
former is not well seen on HRP 630nm SDI. We footnote that an apparently equatorward propagating Tn
intensification structure originating from higher latitudes is indeed seen on the 557.7nm HRP SDI data (not
shown) which observe much lower altitudes. Partly due to the potential uncertainty of Tn measurements,
this study is restricted to qualitative descriptions of certain Tn features in STEVE events, and we shall put
more quantitative analyses on the neutral winds.
The bottom two panels of Figure3 show the east-west and north-south components of the neutral wind
velocity, respectively. Note that in the data processing we have excluded data points at the zenith and the
innermost annulus—by doing so we only keep data points at >25° zenith angle—since the near-overhead
data points are not geometrically suitable for the horizontal wind fitting and thus may contain relatively
large errors. The PKR and HRP SDIs display generally compatible patterns of wind variations, yet certain
discernible differences exist. Some of the discrepancies between the two SDI measurements may be attrib-
uted to the LOS integral effect under an oblique viewing geometry. For example, the winds seen on HRP
SDI near its northern edge of FoV (66°-68° MLAT) are weaker than those seen on PKR SDI. Besides, the
potential existence of longitudinal inhomogeneity (PKR and HRP are separated by ∼3.6° MLON) and the
error of the monostatic fit wind data may also contribute to the differences between the two SDI winds.
Note that the above two error sources are intertwined since the zonal homogeneity of meridional winds
is one of the underlying assumptions of the monostatic fit. Nevertheless, both SDIs show that westward
winds at auroral latitudes start to intensify shortly after the westward plasma flow intensification seen on
PFISR, and such westward wind intensification gradually propagate equatorward. Equatorward winds also
intensify but appear to slightly lag the westward wind intensification. The key observation of our interest
is that, the equatorward wind intensification seems not to penetrate to lower latitudes (≤62° MLAT) before
07:25UT. In particular, the HRP observation clearly shows a peak of the southward wind at ∼64° MLAT,
and an abrupt reduction of the southward winds at ∼63° MLAT starting from ∼07:05UT, ∼15–20min be-
fore the emergence of STEVE. The isolated peak of southward winds is not seen on PKR, yet a diminishing
southward wind toward low latitudes can still be perceived. For simplicity, we shall refer to the latitude
corresponding to a sharp wind deceleration as a “stop latitude” in our presentation, yet we of course do not
mean a complete termination of winds. Such a “stop latitude” of the equatorward winds later becomes the
LIANG ET AL.
10.1029/2020JA028505
7 of 30
Journal of Geophysical Research: Space Physics
initiating latitude around which the STEVE starts to rise. After ∼07:30UT the equatorward winds appear to
penetrate through to further lower latitudes, and the stop latitude also migrates equatorward accordingly.
To better demonstrate the neutral wind pattern preceding the STEVE appearance and surrounding the
STEVE latitude, we present in Figures4 and 5 a series of THEMIS ASI images superposed by the horizontal
neutral wind vectors from PKR (red) and HRP (green) SDIs. We shall call readers' special attention to the
winds surrounding and within the yellow box region overplotted on the images. This box region serves two
purposes. First, it is the region inside which we are to evaluate the wind divergence subsequently. More
importantly, this region contains the lately emerging STEVE (see Figures5c–5d), so that one may readily
check the wind pattern around the STEVE latitude before the actual emergence of STEVE, in the context
of the presumed preconditioning role of the winds. The substorm onset starts at ∼06:18 UT as inferred
from the ASI observations; some small changes of the wind occurred after ∼06:20UT (not shown), but
until 06:37UT (Figure4a) neutral winds remains moderate or mild. They are moderate and mostly directed
north-westward at the auroral latitudes, while mainly eastward at subauroral latitudes in the lower-latitude
portion of the FoV. Westward winds start to enhance and prevail after ∼06:50UT (Figure4b), but the winds
at subauroral latitudes remain weak. Our main research interest is on the wind dynamics after ∼07:05UT.
In a series of successive images after 07:05UT, winds become intensified in westward and southward direc-
tions, and such intensification gradually expands from higher to lower latitudes. However, the wind magni-
tude decreases steeply across the box region. There is a dramatic difference between the winds at latitudes
above and below the box, or equivalently across the later STEVE arc latitude. Even by visual inspection, one
may readily sense that a negative divergence of winds, or equivalently wind convergence, develops in the
box region ∼15–20min before the emergence of STEVE in this region.
The above-presented SDI winds are from the monostatic fit data set. We have also tried the PKR/HRP bistat-
ic approach following the procedures proposed in Anderson etal.(2012a). We remind that LOS velocities
from two stations are inadequate to solve the three components of winds, so that certain assumptions and
geometry limitations are still necessary. More specifically, the derivation of horizon winds from a bistatic
method requires a common observation volume of two radars, and that the observation plane containing
the two LOS directions from each station toward the common-volume data point must deviate substan-
tially from the local vertical direction. This is particularly necessary for our event, since the horizontal
winds in the overlapped area of two SDIs are mostly south-westward directed, roughly perpendicular to
the PKR-HRP line, so that any LOS measurement too close to the PKR-HRP great circle plane would have
a significant yet undetermined contribution from the vertical wind. In the following demonstration, we
adopt the same procedures and geometric criteria to derive the bistatic horizontal winds as those in An-
derson etal.(2012a). Readers are referred to their paper to see the technical details. Figure6 displays four
frames (corresponding to Figures4b, 5a, 5c, and 5d, respectively) of the bistatic winds superimposing on
GAKO ASI images. Due to the geometry limitation, most of the available bistatic data points are relatively
far away from the STEVE arc, so that they are not particularly suitable for our research goal of examining
the neutral wind dynamics surrounding the emerging STEVE arc. That said, the bistatic winds may serve as
a reality check of the monostatic fit data. As one can see, the overall pattern and magnitude of the bistatic
winds are fairly similar to that shown in the monostatic data: (1) winds are intensified during the substorm,
and the intensification propagates equatorward; (2) winds are strong and mostly south-westward at higher
latitudes, yet diminishing toward lower latitudes. We thus claim that the bistatic horizontal winds are gen-
erally consistent with the monostatic data, though upon a close look the bistatic data appear to unveil zonal
inhomogeneity of meridional winds around ∼65° MLAT, which contradicts the underlying assumption,
and constitutes one likely error source, of the monostatic fit.
In the following we shall make quantitative efforts to evaluate the neutral wind divergence surrounding the
STEVE latitude. This is motivated by the notion that the wind divergence alludes to a change of the neutral
density, as per the continuity equation
/ ··dn dt n u
(to be further discussed in Section4). Due to the
potential error in the monostatic data, and the inadequacy of bistatic data points in the vicinity of STEVE,
the two data sets are deemed nonideal for our purpose. Instead, we shall directly use the raw LOS velocity
measurements of HRP and PKR SDIs. The algorithm is conceptually similar to that introduced in Conde
and Smith(1998), but no additional assumption on the wind derivatives is required. For the horizontal com-
ponents of the neutral wind, we use a Taylor expansion approximation to the first-order,
LIANG ET AL.
10.1029/2020JA028505
8 of 30
Journal of Geophysical Research: Space Physics
_0 0 0
··
xx
xx
uu
u u xx yy
xy
(1)
_0 0 0
··
yy
yy
uu
u u xx yy
xy
(2)
in which x (positive north) and y (positive west) denote the meridional and zonal directions, respectively.
A local plane approximation is adopted. x0 and y0 give the center of the region of interest, and ux_0 and uy_0
denote the x and y components of the wind at this center point. The sampling region (i.e., the yellow box
in Figures4 and 5) covers a latitude range of 62.5–64° MLAT that contains the initial rise of STEVE, and a
longitude range spanning from 2.5° west of PKR center meridian to 2.5° east of HRP center meridian. The
SDI measures the LOS velocity component of the vector wind, that is,
LOS
· · ·,
xx yy zz
u ul ul ul
(3)
in which lx, ly, and lz denote the unit vector of the LOS direction in local x-y-z coordinate (z-direction is ver-
tically upward). We assume
_0 _0
, , , /, /, /, /
zx y x x y y
uuu uxuyuxuy
are constant in the box region
of interest. Using an array of LOS measurements contained in the region, we solve the above parameters via
a least squares fitting method.
There is however, one issue in the practical implementation of the above technique, related to the uncer-
tainty of uz and its possible horizontal inhomogeneity. This is because the region of our special interest
contains the overhead data points and includes reducing horizontal winds. uz is typically much smaller
than horizontal winds. In a routine monostatic fit (e.g., Conde & Smith,1998), the LOS wind at the zenith
LIANG ET AL.
10.1029/2020JA028505
9 of 30
Figure 4. A series of ASI images superposed by the monostatic horizontal neutral wind vectors from PKR (red) and HRP (green) 630nm SDIs. An emission
height of 250km is assumed for both ASI and SDI. A yellow dashed box marks our sampling region for wind divergence calculation (see text for details).
Journal of Geophysical Research: Space Physics
data point is often deemed as the uz, which is further assumed to be constant for the entire fitting region.
In our event, uz given by the zenith data point of HRP remains small (<∼20m/s, not shown) and exhibits
some gravity wave-like oscillations all the time. However, if the LOS direction is too close to the zenith,
even a small uz may impose a nontrivial effect on the fitting of horizontal winds. This is the reason why the
near-overhead points in the monostatic fitted wind vector data may contain large errors. As a numerical
test we have tried a few different methods to treat uz: (1) assuming uz constant and using direct least squares
fitting through Equations1 2 3 to obtain uz; (2) using zenith-observed uz in 3 and fitting other parameters;
(3) excluding zenith data point and assuming uz= 0. All three methods yield noticeably different fitting
outcomes of the horizontal gradients if we include all data points with uniform weight. To deal with such a
difficulty, we first exclude the HRP zenith and innermost annulus data, and only keep data points with ze-
nith angle >26° (tan26°∼ 0.5 so that the horizontal wind projection on the LOS direction prevails as long as
it is more than twice the vertical wind). We then adopt a weighted least squares fitting technique: LOS data
points are assigned with different weights, in practice defined by tan(θ), in which θ is the angle between
the LOS and the local vertical direction, so that the data points with LOS direction closer to the local zenith
have a relatively smaller contribution to the overall fitting. The above procedure applies only to HRP SDI;
for the PKR SDI the box region is well off-zenith, and we shall use a uniform weight (taken as the averaged
weight of all HRP data points in use) for PKR data points. Via numerical tests we have found that the above
weighted least squares fitting technique may yield relatively stable horizontal gradients against different uz
assumptions, but the downside is that, admittedly we are not able to reliably derive uz with our procedure.
The top panel of Figure7 shows the measurement uncertainty of the LOS velocity from the Doppler-shifted
630nm emission spectrum for all data points included in our calculation. As one can see, the measurement
uncertainty is always very small in this event. The second panel of Figure7 shows the residual standard er-
ror of the fitting, which is persistently<∼5m/s. The above two errors are small compared to the wind veloc-
ity, so that the fitting quality can be expected to be reasonably good. Based upon the fitted solution of partial
LIANG ET AL.
10.1029/2020JA028505
10 of 30
Figure 5. A continuation of Figure3.
Journal of Geophysical Research: Space Physics
derivatives, we show the divergence (
·u
) and vorticity (
u
)z of the winds in the bottom two panels of
Figure7, respectively. The divergence reaches a negative valley of∼−4×10−4 s−1 during ∼07:10–07:25UT,
preceding the rise of STEVE. This is consistent with our inference from the stop latitude of winds. To fur-
ther corroborate our result, we have also tested with other methods to evaluate the divergence: (1) Using
the monostatic wind vector data of HRP and PKR SDIs (with zenith and innermost annulus data of HRP
excluded); we shall elaborate the involved algorithm in the next subsection. (2) Excluding data at the two
innermost annuli and keeping only data points at >36° zenith angle—by doing so we further suppress the uz
effect but at a cost of reduced data points in the fitting. As one can see, all approaches yield fairly compatible
results. Most importantly, a strong divergence of∼−4×10−4 s−1 before the STEVE emergence is reproduced
in all three approaches. We thus believe that the derived divergence is trustworthy. The above-depicted
technique of weighted least squares fit with the HRP zenith and the innermost annulus data excluded is
deemed successful, and shall be adopted later in the calculation of wind divergence in nonSTEVE events.
The obtained wind convergence in this STEVE event is much stronger than the typical values seen at sub-
auroral latitudes in the evening sector (e.g., Dhadly etal.,2017, 2018; Kwak & Richmond,2014; Thayer &
Killeen,1991). Two footnotes are given here. First, our procedure assumes a uniform divergence over the
entire box region whose latitudinal width (1.5°) is much wider than the STEVE arc. This is of course mainly
limited by the latitudinal resolution of SDI measurements. Second, the convergence in the box region ap-
pears to relax after ∼07:25UT, but this is mostly because that the stop latitude of the winds has gradually
shifted equatorward (and so does STEVE) and moved beyond our calculation region.
The wind vorticity also shows variations before the emergence of STEVE. The enhanced vorticity during
∼07:00–07:10UT is related to the equatorward progression of the westward wind and its shear with the
preexisting eastward wind at lower latitudes, the latter of which is possibly associated with the day-night
temperature gradient in the dusk sector. The vorticity in the box region is reduced as the westward winds
expand further equatorward after ∼07:15UT. While the wind vorticity may be of interest for some other
LIANG ET AL.
10.1029/2020JA028505
11 of 30
Figure 6. A series of ASI images superposed by the bistatic horizontal neutral wind vectors. An emission height of 250km is assumed.
Journal of Geophysical Research: Space Physics
research purposes, in this study, during our event selection procedure we
have noticed that a counterclockwise vorticial wind pattern is fairly com-
mon in SDI data at subauroral latitudes and in the local time range of
interest, and a vorticity of ∼2×10−4 s−1 is often reached or even exceeded
in nonSTEVE substorm events (not shown). As a matter of fact, this lev-
el is basically about the same magnitude as those mean values obtained
from large statistical studies at subauroral latitudes in the evening sector
(Dhadly etal.,2017, 2018; Thayer & Killeen,1991). Therefore, so far we
do not deem the wind vorticity a STEVE-specific feature, and will not pay
much attention to it in our following presentation.
3.2. March 26, 2008 Event
We shall then report another STEVE event that is solely seen on PKR
SDI (The HRP SDI had not launched by that time). This event occurred
during a nonstorm interval (SYM-H >10) but the substorm intensity was
strong (AL reaching ∼800nT). This event is also studied for other re-
search purposes by Nishimura etal.(2020a, 2020c). Based upon available
optical data, the onset of strong aurora intensification was first seen in
the FSIM sector at 06:33UT, which rapidly expanded westward into the
Alaska sector in a few minutes. The top panel of Figure8 shows the keo-
gram of GAKO THEMIS ASI sampled along the GAKO center meridian
assuming a 110km emission altitude. The STEVE of interest is marked.
Though not of the main research objective of this study we bring forth
one note here. Before ∼07:00UT there were some seemingly detached
arcs that reached low latitudes, making one wonder whether they rep-
resent STEVE as well, and what is the difference between them and the
later STEVE of our interest. These previous low-latitude arcs are analyz-
ed by Nishimura etal. (2020a), who found them to be related to strong
energetic proton precipitations as seen in the POES satellite data. STEVE
on the other hand, is known not to be associated with energetic proton precipitation or equivalently proton
auroras (Gallardo-Lacourt etal.,2018a; Liang etal.,2019; MacDonald etal.,2018). A movie is provided to
show the evolution of those previous low-latitude arcs, and some brief descriptions are given in supplemen-
tary material. Via triangulation procedures from multiple ASIs (see supplementary material) we evaluate
the emission altitudes of those previous low-latitude arcs as well as the later STEVE: the former arcs are
found to be centered at an emission altitude of ∼100–110km, while the STEVE is found at ∼250km altitude.
We conclude that those previous low-latitude arcs represent different phenomena from the later STEVE.
The rest of the panels in Figure8 display a series of images from GAKO THEMIS ASI (assuming 250km
emission altitude), showing the STEVE evolution. At around ∼07:22UT. the STEVE arc begins to emerge
in the eastern portion of the ASI FoV. It expands progressively westward in the next a few minutes, and
reaches the FoV of PKR SDI by ∼07:25UT. Later on, the STEVE becomes more brightened and dynamic,
with some signs of co-existing double-layer STEVE and/or “picket-fence” structures (Liang etal., 2019),
but a detailed analysis of the STEVE structures is beyond the interest of this study. A supplementary movie
displays the evolution of the STEVE. This STEVE and its later picket-fence structure were captured by pho-
tographs from a citizen scientist, Mr. Challak, at Palmer, Alaska (see FigureS3). Nishimura etal.(2020a)
found that, DMSP F16 satellite data, though in the southern hemisphere, crossed a strong SAID channel
that was roughly conjugate to the northern STEVE arc in this event.
The top panel of Figure9 shows the keogram sampled along the center meridian of PKR SDI, from GAKO
THEMIS ASI data. The second panel shows the east-west convection velocity deduced from PFISR observa-
tions. Enhanced westward plasma flows of order of ∼1km/s are seen near the equatorward edge of its FoV.
The third panel shows the neutral temperature keogram derived from PKR SDI. There is a Tn enhancement
that originates from auroral latitudes at ∼66° MLAT after ∼06:45UT, and there is another Tn enhancement
LIANG ET AL.
10.1029/2020JA028505
12 of 30
Figure 7. The top panel shows the residual standard error of the least
squares-fitting in our calculation of the wind divergence. The middle and
the bottom panel shows the calculated divergence and the z-component of
vorticity of horizontal neutral winds. For the wind divergence of our key
interest, we have compared with the results from two other approaches:
the dashed curve denote the approach using the monostatic wind vector
data of HRP and PKR 630nm SDIs, while the dotted curve denote the
approach that excludes data at the two innermost annuli and keeping only
data points at >36° zenith angle. See text for details.
Journal of Geophysical Research: Space Physics
at latitudes <63° MLAT, which appears to stem from the low-latitude edge of the SDI FoV after ∼06:50UT.
The Tn observation is thus consistent with that in the previous event, except that the enhancements at both
higher and lower latitudes are seen in one SDI instead of being seen via two separate SDIs as in the previous
event. In this event, the Tn intensification at lower latitudes is strong enough to survive the LOS integral
effect at low-elevation measurements. The fourth and fifth panels of Figure9 show the east-west and north-
south components of the neutral winds. Westward winds start to enhance after ∼06:45UT, initially at ∼66°-
68° MLAT, and then gradually propagate toward lower latitudes. The southward winds also intensify, but
LIANG ET AL.
10.1029/2020JA028505
13 of 30
Figure 8. The top panel shows the keogram of GAKO ASI based on 110km emission height. The rest of panels display a series of GAKO ASI images based
on 250km emission height, showing the evolution and westward propagation of STEVE for the March 26, 2008 event. The location of PRK SDI is marked as
triangle.
Journal of Geophysical Research: Space Physics
we again notice that the southward neutral wind seems not to penetrate to lower latitudes; there is a distinct
reversal from southward winds to northward winds across ∼63° MLAT. Such a wind reversal latitude is
evident after ∼07:12UT. The STEVE later initiates at roughly the same latitude as this reversal latitude of
meridional winds.
We also evaluate the wind divergence around the STEVE latitude. We shall first clarify that it is not mathe-
matically possible to fit the wind gradients using single-station LOS measurements with a procedure similar
to that in the previous subsection, without additional assumption on the wind derivatives (for mathematical
LIANG ET AL.
10.1029/2020JA028505
14 of 30
Figure 9. The top panel shows the ASI keogram sampled along the center meridian of PKR. The second panel shows
the east-west convection velocities (positive west) from PFISR measurements. The third panel shows the neutral
temperature measurements from PKR 630nm SDI. The fourth and fifth panels of each subfigure show the east-west
and north-south components of the neutral wind velocity observed by PKR SDI. From the third to fifth panel, a black
dotted curve marks the trace of the STEVE arc. The bottom panel shows the calculated wind divergence at 62.5°-64°
MLAT.
Journal of Geophysical Research: Space Physics
details in this regard, see Conde and Smith[1998]). Instead, we shall directly use the monostatic-fitted hori-
zontal wind vectors to evaluate the divergence. We sample the vector wind data within 62.5°-64° MLAT and
±3° MLON around the PKR center meridian. The data points in the sampling region are ∼42°-52° off the
PKR zenith so that the uz uncertainty is trivial there. We apply a first-order Taylor expansion of the vector
winds, also defined by Equations1 and 2, in the region of interest, and evaluate wind derivatives using a
least squares linear regression. The outcome of the computed divergence is plotted in the bottom panel of
Figure9. The wind convergence starts to enhance after ∼07:12UT, reaching ∼4.5×10−4 s−1 before the emer-
gence of STEVE. The above observations are largely consistent with that in the previous event.
3.3. Summary of Two STEVE Events
In the above we have described the neutral wind and temperature variations in two STEVE events. To
summarize the key observations, neutral winds enhance in westward and southward directions following
substorm auroral intensification, and show southward propagating trend originating from auroral latitudes.
However, the equatorward winds appear to feature a steep stop/reversal at certain latitude, and strong neg-
ative divergence of wind develops there. This pattern sustains for ∼15–20min, and then STEVE arises at
about this stop/reversal latitude. Neutral temperature enhancements take place in two latitude ranges: one
originating from auroral latitudes, and the other stems from latitudes lower than STEVE.
There are several existing studies on neutral wind variations at high latitudes during substorm intervals,
some of which showed conflicting findings (e.g. Cai et al., 2019; Ritter et al., 2010; Xu etal.,2019; Zou
etal.,2018, 2020). Ritter etal.(2010) and Xu etal.(2019) both reported fairly small wind variations (no more
than a few tens m/s) after substorm onset, and the deviation tends to be eastward in the premidnight sector,
yet Cai etal.(2019) and Zou etal.(2020) both noticed the presence of much larger (>100m/s) variation
magnitudes with westward/southward wind intensification. One of the possible reasons for the above dis-
crepancy might be the difference in the observation latitudes relative to substorm-intensified auroras. In the
premidnight sector, plasma convection typically follows a Harang system, eastward at higher latitudes and
westward at lower latitudes. In particular, flows near the equatorward boundary of auroras often strong-
ly enhance with the auroral intensification and become part of the SAPS (Lyons etal., 2015; Nishimura
etal., 2008; Zou etal.,2009; 2012; Gallardo-Lacourt et al., 2017; Zou et al.,2018), as partly corroborated
by PFISR data in our events. Westward neutral winds in the F-region ionosphere intensify accordingly due
to the ion drag, and the neutral temperature enhances due to a combination of the auroral heating and
the Joule heating associated with the strong convection electric field. This may generate an equatorward
pressure gradient force toward subauroral latitudes which leads to the southward wind. It is important to
note that the meridional and zonal winds are inherently coupled via the Coriolis force. Thus, the southward
pressure gradient force may also partially contribute to the westward rotation of winds due to the Coriolis
effect. However, in a statistical study Zou etal.(2020) found that a southward wind intensification is more
often seen in the midnight-postmidnight sectors, but not common in the evening-premidnight sector. To
the authors' knowledge, none of the existing observations of substorm-associated neutral winds specifically
addressed the wind divergence at subauroral latitudes. It is mandatory for us to investigate whether or not
the subauroral neutral wind pattern we observed during STEVE events is common to all substorm events,
which is the goal of the next subsection.
In both STEVE events, neutral temperature enhancements take place in two latitude ranges: one originates
from auroral latitudes, and the other is at latitudes lower than the STEVE and appears to stem from the
low-latitude edge of the SDI FoV. This may not be a coincidence. As stated above, the equatorward wind
is mainly driven by the pressure gradient force that is associated with the Joule heating in the auroral and
SAPS flow regions. If in the meantime there is also a strong temperature enhancement at mid-latitudes
and in turn an equatorward temperature gradient, the corresponding pressure gradient force (assuming
the density gradient is comparatively minor) would act to impede the equatorward winds, thus explaining
their deceleration/stop at certain latitude. The equatorward pressure gradient may even lead to poleward
winds at lower latitudes, as hinted in March 26, 2008 event (see Figure9). The origin of such a Tn intensifi-
cation probably owes its source in the mid-latitude region, beyond the SDI FoV. While this is an interesting
research topic by itself, we realize that mid-latitude Tn enhancements may own their source to various
LIANG ET AL.
10.1029/2020JA028505
15 of 30
Journal of Geophysical Research: Space Physics
mechanisms, some of which may not be related to substorm processes at all. As we indeed infer from our
survey, mid-latitude Tn enhancements at times exist in nonSTEVE events as well. Due to the lack of rele-
vant observations at mid-latitudes in this study, it is difficult to clarify the source mechanisms of those Tn
enhancements and make a meaningful comparison between STEVE and nonSTEVE events. This is further
confounded by the fact that the temperature measurements usually contain larger uncertainties than the
wind velocity measurements (e.g., Andersen etal.,2012a), particularly at lower elevation angles where a
LOS integral effect may be detrimental. Due to the above reasons, in the following we shall focus on the
neutral winds in our analyses of nonSTEVE events.
3.4. Comparison with NonSTEVE Substorm Events
For a comparison purpose, we survey over the year 2010–2012 to collect nonSTEVE substorm events, and
investigate their general pattern of neutral winds at subauroral latitudes. Our event selection criteria are as
follows.
1. The neutral wind patterns, both the background one and the substorm-associated variations, have a
strong local time dependence (e.g., Emmert etal., 2006; Dhadly et al., 2017, 2018). In particular, Zou
etal. (2020) indicated a systematic difference between the neutral wind pattern in the evening sector
and that in the midnight sector during substorm intervals. For both STEVE events we presented above,
the substorm onset occurs during 06:00–07:00UT, while STEVE mainly exists during ∼07:25–08:00UT
or ∼20.5–21.1h MLT. To screen the influence of the local-time dependence of neutral winds on our
comparison effort, we shall only look for events with substorm onset at 6–8 UT, and accordingly mainly
investigate neutral wind variations during ∼07:00–09:00UT. Neutral wind variations at other time sec-
tors and comparison for STEVE and nonSTEVE events will be left for future studies
2. The minimum AL index must be less than −200nT for the substorm to be considered. To define a sub-
storm onset, we use magnetometer and optical data in Alaska and western Canada sectors. While it is
often the case that not all ASIs have favorable viewing conditions, clues of auroral intensification and
expansion must be seen in at least one optical ASI in Alaska, so that we can reasonably conceive that
the substorm activity has indeed approached the time sector of SDI measurements. If the optical view-
ing condition is ideal so that the initial onset time and location can be unambiguously identified from
ASIs, we shall use this optical onset time. Otherwise, if optical observations in the initial onset sector
are unavailable, we shall resort to the SuperMAG database for the onset time (Gjerloev,2012). This mag-
netometer-based onset time may lag the actual onset time by several minutes, but such uncertainty is
trivial to this study, since the onset time is used in this study only to delimit an overall time interval for
us to look into the SDI data
3. The viewing condition of GAKO THEMIS ASI (or equivalently the digital all-sky-camera at GAKO run
by the University of Alaska) during the event interval must be good enough for us to exclude the exist-
ence of STEVE. While it is possible to derive the STEVE-related neutral dynamics with PKR SDI alone,
as we have exemplified in Section3.2, a combination of HRP and PKR data is preferable. Thus, in the
survey of this study we preferentially select events with SDI data available at both HRP and PKR, though
later in this subsection we shall also enroll and analyze some PKR-only events from Zou etal.(2020)'s
database to complement this study. Since GAKO is collocated with the HRP SDI, the criterion that the
former has a good viewing condition naturally means that the latter has as well, but we make additional
checks (via the intensity map) on the PKR 630nm SDI data to make sure the sky is not cloudy, at least at
southward looking directions, that is, toward subauroral latitudes of interest
4. Our research interest is on the STEVE and neutral wind dynamics at subauroral latitudes, and we resort
to GAKO ASI and HRP SDI for such a goal. If the auroras expand to too low latitudes and populate the
SDI FoV, neutral wind variations seen by HRP SDI would be more pertinent to auroral dynamics than to
subauroral processes, and STEVE, even if existing, could be out of the FoV of GAKO ASI. Therefore, for
the above technical reasons we exclude events when the main auroral oval is driven to, and persistently
(>30min) stays at, very low latitudes (<62° MLAT). Such a criterion bars some intense storm/substorm
events in which auroras expand substantially equatorward, as we shall address later in this section
5. Gallardo-Lacourt et al. (2018b) found that STEVE may have a longitudinal extent of ∼2,145 km, or
∼2.5h MLT assuming an average latitude of ∼60°. A similar conclusion was also reached in Nishimura
LIANG ET AL.
10.1029/2020JA028505
16 of 30
Journal of Geophysical Research: Space Physics
etal.(2020c). Since the available wind data are from PKR/HRP SDIs (∼−90° MLON), our search for the
optical signature of STEVE is limited to ASIs in Alaska and western Canadian sectors, with Athabasca
(∼−52° MLON, see Figure1) the eastern-most ASI station to be considered. It is possible that STEVE
may exist in further eastern sectors (e.g., February 8, 2016 event to be presented later in this section),
but such a case may not invalidate our research goal, namely the possible effects of neutral winds on the
local existence/nonexistence of STEVE
Over 2010–2012, we have identified 10 events that meet the above criteria. Their neutral wind variations
observed by HRP SDI from substorm onset to 1.5h after are presented in Figures10–12. This time range
is based upon Gallardo-Lacourt etal.(2018b)'s finding that, the emergence of STEVE usually lags the sub-
storm onset by ∼50–80min. Any neutral wind variations that might contribute to the rise of STEVE should
presumably transpire and sustain for a while before the latter, and thus should have occurred no much
beyond ∼1h after the onset. In most of the events, the temporal resolution of the SDI measurements is
∼2–4min which suffices for our research purpose. There are a few events (event F/G, and partly in event
I/J) in which the SDI at times runs in a low-resolution mode, so that large gaps (∼13–14min) may exist
among the data. Such data gaps do not significantly affect our evaluation of the overall wind variation
pattern after substorm onset, but the wind divergence calculation during these gaps would be missing,
adding uncertainty to our evaluation of the maximum wind convergence (see later in this subsection and
Figure14d) in those events. The minimum AL in each event reached in each event ranges between∼−250
and −700nT, weaker than the two STEVE events reported abut not systematically different from the overall
AL range of STEVE occurrence (Gallardo-Lacourt etal.,2018b). An intense substorm (AL∼−950nT) from
Zou etal.(2020)'s events will be separately introduced later in this subsection.
Each subfigure in Figures10–12 denotes one event, which is indexed to facilitate reference. For each event
subfigure, the top panel gives the auroral keogram from the GAKO THEMIS ASI. In the second and third
panels we plot the keogram of the east-west wind and the north-south wind from HRP SDI data using the
same procedure as for the STEVE events. The neutral wind velocities in all events are plotted in the same
color scheme as that in the two STEVE events, to facilitate a cross-comparison. On top of each event subfig-
ure we also show the minimum AL index reached during the substorm, as a proxy of the substorm strength.
In the bottom panel of each subfigure we show the wind divergence within two latitudinal ranges: ∼61°-
62.5° (green) and 62.5°-64° MLAT (black), calculated from the procedure introduced in Section3.1. The lat-
itude width used in the divergence calculation is the same as that for the two STEVE events, designed to ac-
commodate enough (>10 as our criteria) data points to ensure numerical reliability in a least squares fitting
against instrumental noise and occasionally bad data points. The PKR SDI data are not presented here but
are involved in the calculation of the wind divergence. Note that Event F has no PKR data available, but in
this event the winds are dominantly northward. The pattern is outright opposed to that in STEVE cases, so
that the wind convergence is not important for our arguments and thus not calculated there. We emphasize
again that we intend to compare the subauroral neutral wind pattern during nonSTEVE events with that
during STEVE events. Occasionally in some events, there may be some hints of wind divergence at >64°
MLAT (e.g., event H/J, and see Figure13 later), but they occur at auroral latitudes (see top panel) and are
not much relevant to our research interest. STEVE is never found to occur at >64° MLAT to the authors' rec-
ognition, while latitudes at <61° MLAT contain insufficient data points for a reliable computation of wind
divergence from our algorithm. Therefore, the latitude ranges in which we calculate the wind divergence
is delimited between 61° and 64° MLAT. Such a range covers the latitudes of initial STEVE emergence in
many realistic STEVE events according to the existing literature (e.g., Gallardo-Lacourt etal.[2018b]) and
our experience. It is true that STEVE might at times occur at even lower latitudes (e.g., <60° MALT), but
identification of the existence/nonexistence of those lower-latitude STEVEs and their possible relationship
with neutral winds is beyond the instrumental and methodological capability of this study.
A westward wind enhancement after substorm onset is seen in most but not all nonSTEVE events. Since the
local time sector of interest is close to the dusk, an eastward wind led by the day-night temperature differ-
ence is expected and indeed often seen, particularly at lower latitudes. The westward wind intensification
is presumably contingent upon the strength of the substorm-related SAPS and its competence with the
ambient eastward wind. The meridional wind is more of interest in this study. In some of the events (event
B/C/F/I), mainly northward winds mixed with some sporadic weak equatorward winds prevail in ∼1h after
LIANG ET AL.
10.1029/2020JA028505
17 of 30
Journal of Geophysical Research: Space Physics
LIANG ET AL.
10.1029/2020JA028505
18 of 30
Figure 10. Each subfigure represents one nonSTEVE substorm event. In each subfigure, the top panel shows the ASI keogram sampled along the center
meridian of HRP based on 110km emission height. The second and third panels are the keogram of the east-west wind and the north-south wind from HRP
630nm SDI. The bottom panel shows the calculated wind divergence at 62.5°-64° MLAT (black) and 61°-62.5° MLAT (green). The substorm onset time is
marked by a vertical dashed line.
Journal of Geophysical Research: Space Physics
onset. This is compatible with Zou etal.(2020)'s result that, a southward wind intensification after substorm
onset is not found as statistically prevailing in the evening sector. In those events with perceptible south-
ward wind intensification, except for few events (event D/J) the intensifications at the transition latitude
(∼64°-65° MLAT) from auroral to subauroral regions tend to be about ∼25%–50% weaker than that achieved
in the STEVE events. At subauroral latitudes (<∼64° MLAT) of interest, the southward winds in general
appear to extend well into low latitudes. There are occasionally singular gaps/blanks among the data due to
bad/missing data points, but there is little sign of systematic and long-standing (>10min) manifestation of
“stop latitude” as in STEVE events. The wind convergence calculated in the 61°-62.5° and 62.5°-64° MLAT
ranges seldom exceeds ∼2×10−4 s−1. One event that somehow resembles the STEVE events in terms of neu-
tral wind variations is event J. The southward wind intensification is strong—we however footnote that, as
compared to the two STEVE events the local time of observation in event J is closer to midnight, where both
the ambient trend and substorm-associated components of the southward winds are supposedly stronger
(Emmert etal.,2006; Zou etal.,2020). There appears to be a gradually decreasing trend of southward winds
LIANG ET AL.
10.1029/2020JA028505
21 of 30
Figure 13. The top panel shows the 630nm auroras observed by PKR MSP on February 8, 2016 based on 230km emission height. The second and third panels
show the east-west and north-south components of the neutral wind velocity observed by PKR 630nm SDI. The bottom panel shows the calculated wind
divergence at 62.5°-64° MLAT.
Journal of Geophysical Research: Space Physics
toward lower latitudes during ∼08:30–08:50UT in this event, but the de-
creasing slope is less abrupt and, the wind convergence is weaker there
(up to ∼2.4×10−4 s−1) than in STEVE events.
Figure12k shows the averaged keogram of zonal and meridional winds,
as well as the averaged wind divergence in the two subauroral latitude
ranges, for the 10 collected events. In the data processing we have av-
eraged/interpolated the data to ten 10-min bins spanning from −10 to
90min relative to onset. There is a perceptible trend of westward wind
intensification after the substorm onset, which subsequently progresses
southward. There also seems to be a subtle enhancement of southward
winds at auroral latitudes after the onset. We however remind that an
enhancement toward ∼90min after the onset, when the substorm itself
has subsidized in most of our events, could be at least partly attributed
to an ambient diurnal trend, namely that neutral winds tend to become
more southward when approaching local midnight (Emmert etal.,2006;
Zou etal.,2020). At subauroral latitudes, the meridional winds are weak
and seem to be more or less constant or slowly vary across latitudes, and
the wind divergence there is tiny or even positive on average.
Our analyses are supplemented by the premidnight event pool in Zou
et al. (2020), who independently surveyed substorm events with SDI
measurements from late 2012 to 2017, though their main research inter-
est is at auroral latitudes. They also stipulate that AL< −200 nT must
be reached for the substorms chosen, but they did not check optical data
and did not distinguish STEVE or nonSTEVE events. Most of their events
(13 out of 15) have only PKR SDI observations, without GAKO THEMIS
ASI and HRP SDI data, since the latter two instruments both ceased op-
eration in early 2014. However, we have checked other available THEMIS
ASI observations in Alaska and western Canada sectors (up to Athabasca
ASI, see afore-mentioned event criterion 5) found no evidence of STE-
VE in available ASI data for those events. Also, digital all-sky-camera
data at GAKO run by the University of Alaska at Fairbanks are available
in five of Zou etal.'s events after 2016, and in none of them any visible
signature of STEVE is found. While Zou etal.'s events are not ideal in
distinguishing the existence or nonexistence of STEVE, they may never-
theless be useful in providing complementary information on the wind
convergence at subauroral latitudes during substorm intervals as well as
its possible dependence on the substorm intensity. Upon checking availa-
ble optical data we exclude February 17, 2017 event, in which the auroras
are found to extend to very low latitudes (∼62° MLAT), in the following
analyses as per our criterion (4).
We exemplify one such PKR-only event in Figure 13. This February 8, 2016 event represents an intense
substorm: its AL reaches ∼950nT which is even stronger than the two reported STEVE events. Note that in
this event STEVE was seen on the Lucky Lake (∼59.1° MLAT, −42.5° MLON) red-line imager (data availa-
ble at https://data.phys.ucalgary.ca/sort_by_project/GO-Canada/), which is more than 3h MLT east and 4°
MLAT south to the GAKO station, after ∼06:45UT. However, no STEVE is visible in the observations from
the GAKO digital all-sky-camera (not shown, data available at ftp://optics.gi.alaska.edu/GAK/DASC/) and
the Whitehorse (WHIT) THEMIS ASI. Even if the STEVE did extend over ∼3h MLT into Alaska sector, it
was likely situated at lower latitudes beyond the FoV of GAKO/WHIT cameras and PKR SDI. With such
a possibility in mind, the event may be better depicted as a “local nonSTEVE event,” within the latitude/
MLT scope of available observations. That said, we argue that the event may still be suitable for our research
objective to study the potential role of neutral winds on the local existence/nonexistence of STEVE. The top
panel of Figure13 shows the 630nm emissions derived from PKR MSP observations. The second and third
LIANG ET AL.
10.1029/2020JA028505
22 of 30
Figure 14. (a) Calculated wind divergence in 61°-62.5° MLAT,
superimposed according to the time relative to substorm onset from 13
HRP+PKR events. (b) same as (a) but for 62.5°-64° MLAT. (c) same as (b)
but for 13 PKR-only events. The two STEVE events are highlighted in color
in (b and c). (d) The minimum wind divergence reached in each event,
versus the mean AL averaged in 0–30min before the minimum divergence
epoch. Asterisks and diamonds denote HRP+PKR events and PKR-only
events, respectively.
Journal of Geophysical Research: Space Physics
panels show the zonal wind and the meridional wind observed by PKR SDI. Accompanying each auroral
equatorward expansion at ∼05:50, ∼06:25, and ∼07:00 UT, there is an equatorward propagating trend of
intense westward winds. The southward winds are however, not as strongly intensified. They remain very
weak during the first auroral expansion, and are moderately intensified for the second auroral expansion,
but the enhanced southward winds are essentially confined to >64° MLAT, that is, mostly within auroral
latitudes. They seem not to penetrate well into subauroral latitudes. We however cannot entirely rule out the
possibility of the LOS integral effect as one contributing cause of the diminishing winds at lower latitudes,
which is a potential issue for the PKR-only events. For the third and strongest auroral intensification and
expansion after 06:50UT, the southward winds enhanced accordingly and penetrated down to the equator-
ward edge of the PKR SDI FoV, yet without any evident sign of a stop latitude. To summarize the observa-
tions, the southward winds either do not appear to enter subauroral latitudes, or do not have a distinct stop
latitude in the subauroral region. The bottom panel of Figure13 shows the wind divergence calculated in
the latitude range 62.5°-64° MLAT using the procedure described in Section3.2. As one can see, though this
substorm is stronger, the wind divergence is noticeably weaker than that in the STEVE events.
We calculate the wind divergence for all Zou etal.(2020)'s premidnight events. For the two events with
PKR and HRP SDI data, we apply the algorithm described in Section3.1 in two latitudinal ranges: 61°-62.5°
and 62.5°-64° MLAT. Together with April 4, 2010 STEVE event and nine nonSTEVE events from the survey
in this study (event F has no PKR SDI data), there are 12 HRP+PKR events in total. The other 12 events
from Zou etal.(2020) only have PKR SDI data; we adopt the method described in Section3.2. Note that for
those PKR-only events we can only calculate the divergence in 62.5°-64° MLAT, since the 61°-62.5° latitude
range has insufficient (three within ±3° MLON of the PKR center meridian) PKR data points for a reliable
fitting. Figures14a–14c give the plots of the calculated wind divergence superimposed according to the time
relative to the substorm onset, from 12 HRP+PKR events and 13 PKR-only events (including the March 26,
2008 STEVE event), respectively. The two STEVE events (both occurring at ∼63° MLAT) are highlighted in
color in Figures14b and14c. It can be seen that, in STEVE events the wind convergence reaches outlier val-
ues that are rare in any other substorms we have investigated. This indicates that a strong wind convergence
is not a common feature in normal substorm events.
One may wonder that the STEVE events feature stronger wind convergence simply because their underly-
ing substorm strength is stronger than that in nonSTEVE events. A counterexample in this regard has been
given in Figure13. To further investigate this, we identify the maximum convergence (in terms of average
over sliding 10-min windows) reached in each event, and then calculate the mean AL averaged over the in-
terval 0–30min before the maximum convergence epoch. In doing so we have taken into consideration the
typical ion-neutral coupling timescale (Nishimura etal.,2020b). For the HRP+PKR event, the maximum
convergence is determined as the larger one found in the two latitude ranges, while for the PKR-only events
the convergence is only calculated in the 62.5°-64° MLAT range. The outcome is shown in Figure14d.
Results from two STEVE events are highlighted in color. There might be a vague trend of increasing wind
convergence with the substorm intensity, but such a trend is slim to say the best. For the nonSTEVE events,
the maximum convergence is no more than ∼2×10−4 s−1, regardless of the substorm intensity, while in
STEVE events the wind convergence stands out with outlier values. However, we admit that there is a rela-
tive scarcity of intense substorms, particularly in the event pool collected in this paper with HRP SDI data.
One of the reasons for the lack of intense substorm events is due to our criteria (4): since we aim to study
subauroral wind dynamics with HRP SDI, we have ruled out events when a main portion of the SDI FoV is
immersed in auroras, which is often the case under very strong substorms. Based upon available events, we
are unable to firmly answer the inquiry that, whether the wind convergence at subauroral latitudes would
tend to occur more often and be stronger when the substorm is sufficiently intense. This shall be left for fu-
ture studies, ideally with SDI/ASI measurements at more equatorward latitudes. We footnote that a related
question that whether the STEVE occurrence is statistically biased toward more intense substorms is not
firmly answered to date either.
To summarize, from a survey of nonSTEVE substorm events and a comparison with STEVE events, we find
that the southward wind intensifications tend to be in general weaker in nonSTEVE events. In some of the
nonSTEVE events, the winds at subauroral latitudes remain mainly northward. For the rest of events with
southward wind intensification after substorm onset, there is in general no well-defined signature of a stop
LIANG ET AL.
10.1029/2020JA028505
23 of 30
Journal of Geophysical Research: Space Physics
latitude of southward winds at subauroral latitudes, and the wind convergence calculated at subauroral
latitudes is found to be consistently weaker than that in STEVE events.
4. Discussions
In this study, we investigate the patterns and variations of the neutral wind and temperature from the sub-
storm onset to the STEVE emergence. Some of the observed variations can be understood in the context
of the substorm auroral intensification and the subsequent development of SAPS flows equatorward of
the intensified auroras as a part of the Harang system (e.g., Nishimura etal.,2008; Zou etal.,2009, 2012).
The SAPS flows impose two important influences on the neutral winds: (1) the intensification of westward
winds via ion drag and (2) the neutral temperature enhancement via Joule heating. An equatorward wind
may arise due to a combined effect of ion drag led by equatorward plasma flows and a poleward pressure
gradient associated with the SAPS-related Joule heating. The westward and southward wind intensifica-
tions are evident and relatively strong in STEVE events. However, in a survey Zou etal.(2020) found that
southward wind intensifications are not statistically prevailing in the premidnight sector during normal
substorms (they did not distinguish STEVE and nonSTEVE events). We also notice from our survey that,
the winds are dominantly northward in some nonSTEVE events and are only moderately enhanced in some
other events. It thus appears that the southward wind intensification tends to be more pronounced in STE-
VE events than in a majority of nonSTEVE events. While the overall substorm strength is likely one poten-
tial factor controlling the thermospheric response, one other possible reason underlying the different wind
variations during STEVE and nonSTEVE events might be the duskward extension of the magnetospheric
energy input. In a recent study Nishimura etal.(2020c) suggested that magnetosphere injection and con-
vection configurations are more skewed to premidnight during STEVE intervals. For the local time range
under investigation in this study, the GAKO sector is close to dusk. The substorm disturbance usually origi-
nates east of Alaska and then propagated westward. In this regard we have checked the magnetometer and
auroral (mosaic movie available at https://data-portal.phys.ucalgary.ca/themis/mosaic_movies) data and
notice that, during both two STEVE events reported in this paper, the substorm bulge well extends to Alas-
ka before STEVE emergence, and thus the level of magnetospheric energy input in the STEVE region was
presumably larger. On the other hand, in some of the nonSTEVE events, while the substorm activity does
reach Alaska, that is, in the form of arc intensifications, a full-fledged substorm bulge is not always extended
to and well established in the GAKO sector. A larger magnetospheric energy input at premidnight due to an
extended bulge activity for the STEVE events may potentially lead to a stronger thermosphere response in
the STEVE occurrence region (Nishimura etal.,2020c).
Another interesting finding in this study lies in that, the intensified equatorward winds in STEVE events
appear to feature a stop/reversal at certain latitude, and strong wind convergence (∼4–5×10−4 s−1) is de-
veloped there. This pattern sustains for ∼15–20min, and then STEVE arises at about this stop latitude. We
remind that, to ensure the numerical reliability in a least squares fitting against the instrumental noise and
the uncertainty led by uz, the wind divergence is computed over a latitude width (1.5° MLAT) much wider
than the STEVE arc (0.1°-0.2° MLAT). In comparison, an evident stop latitude of southward winds is not
as well seen, and the calculated convergence at subauroral latitudes is consistently smaller, in nonSTEVE
events. From the achieved results in this study we are able to claim that a strong wind convergence is not a
common feature during normal substorm intervals at subauroral latitudes in the evening sector of interest.
That said, since there are inadequate intense substorms in our event pool, we are unable to determine in
this study whether the wind convergence would occur more often and be stronger when the substorm is
sufficiently intense. Anyway, we emphasize that we only hypothesize the strong wind convergence as one
contributing factor, not a sufficient condition, for STEVE to occur, as we shall elucidate later in this section.
In the following, we shall discuss the potential implication of our findings to the possible mechanism of
STEVE production. One of the most striking features of STEVE is that it contains an overall enhancement
of a continuous spectrum as its main source of brightness (Gillies etal.,2019; Liang etal.,2019). The most
likely source of such AGC is a certain kind of chemiluminescence. Several potential mechanisms of the
AGC are being investigated by the authors and colleagues, but one of the leading difficulties in these inter-
pretation efforts lies in that, the observed optical intensity of STEVE summons a high concentration of rel-
evant neutral constituents (including their excited states) involved in the chemiluminescence. In contrast,
LIANG ET AL.
10.1029/2020JA028505
24 of 30
Journal of Geophysical Research: Space Physics
many of the key relevant constituents (such as NO, N (2D, 4S), and excited N2 in a NO2 continuum scenario)
are supposed to have a low concentration in the nightside subauroral region in absence of electron precip-
itation. We propose that the transport effects led by neutral winds might partly help solve this dilemma.
More specifically in terms of our observations: (1) Enhanced equatorward neutral winds can transport rel-
evant neutrals species, which are excited by auroral precipitations and may have crucial importance in the
airglow production, to subauroral latitudes (Solomon etal.,1999; Vlasov & Kelley,2003). (2) Furthermore,
the neutrals may pile up due to the wind convergence, potentially leading to an enhancement of their
densities. It should be noted that the gas continuity equation is inherently coupled with the momentum
equation, such that the relationship between the neutral density variation and the wind divergence is com-
plex. Dhadly etal.(2015) suggested that the thermosphere is largely convectively stable. Their argument is
based on the consideration that the density and pressure changes led by the wind divergence would oppose
the original divergence/convergence that produced them. While this may be true for the major neutral
species (e.g., atomic oxygen in the upper thermosphere), we argue that minor neutral species (including
excited-state neutrals) would be more passive to the neutral wind variations set up by external pressure
gradients, since those constituents have minor only contributions to the total pressure. There is also a pos-
sibility that the horizontal wind divergence may be balanced by the vertical transport term
·nz
u
. In the
afore-addressed context that westward/southward wind enhancement and Tn intensification originating
from auroral latitudes are related to SAPS and its resulting Joule heating, one would also expect a substan-
tial neutral upwelling, which acts to lift the molecule-rich air from lower to higher altitudes and change
the neutral composition there (e.g., Wang etal.,2012; Zhang etal.,2014). SDI observations cannot yield the
distribution and height profile of the vertical wind. In April 4, 2010 event, uz given by the HRP zenith data
point is small (<20m/s, not shown) and mostly downward during the interval of strong horizontal wind
convergence. We have also tried the bistatic method in Anderson etal.(2012b) to derive uz for this event.
While unfortunately the common-volume locations suitable for the Anderson etal.(2012b) uz analysis (see
their Figure1) are all at≥∼64° MLAT and thus not in the subauroral region of our core interest, the derived
uz pattern at those locations (not shown) are found to be semi-quantitatively similar to that inferred from
the HRP zenith data point. If such a pattern is assumed as the general pattern of uz at the SDI observation
height (∼250km), one may imagine that the upwelling from the lower thermosphere might lead to the con-
vergence of molecular air at certain altitudes. Though the lack of definite uz observations inevitably brings
uncertainty, we argue that there seems to be no straightforward rationale to conceive that the vertical trans-
port term
·nz
u
would necessarily act to cancel, if not to reinforce, the convergence of some molecular air
constituents. Combining the above thoughts, it is not unreasonable to conceive that at least some neutral
constituents might undergo substantial pileup due to the observed wind convergence.
Of course, we are not to claim that the neutral pileup led by neutral winds may itself be sufficient in solving
the afore-mentioned difficulty of chemiluminescence airglow—it certainly cannot. To account for the dras-
tic increase of the STEVE airglow brightness, an external energy source must be involved. Based upon the
current knowledge of STEVE, the SAID becomes the topmost candidate for such an energy source. SAID
is similar to SAPS in a number of aspects (some researchers consider SAID as a subclass of SAPS), but is
usually characterized by somehow stronger flow magnitude and narrower latitudinal width (Anderson &
Landry,2017; Mishin etal.,2017). Elevated electron temperature and depleted electron density are typically
observed within the SAID channel (e.g., Archer etal.,2018, 2019a; Moffett etal., 1998; Spiro etal.,1979).
STEVE is found to be co-located with such a SAID channel in joint optical and in situ satellite observa-
tions (Archer etal., 2019a; Chu et al., 2019; MacDonald et al.,2018; Nishimura etal., 2019). Nishimura
etal.(2020c) suggested that substorms and particle injection extending far duskward away from midnight
offer a condition for creating SAID and STEVE due to stronger electron injection to premidnight. Some of
the STEVE features, such as the narrow latitudinal width and especially its fast westward propagation, are
likely pertinent to SAID characteristics. In both of our STEVE events, the STEVE shows a westward expan-
sion: it appears first near the eastern edge of the GAKO ASI FoV and then progressively expands westward.
The westward expansion speed is evaluated as ∼3–5km/s, which is compatible with the typical STEVE-re-
lated SAID velocity (Archer etal.,2019a).
The exceptionally high ion kinetic energy and electron temperature characteristic of SAID open new possi-
bilities of chemiluminescence airglow processes. We shall use the NO2 continuum (
2
NO O NO hν
)
LIANG ET AL.
10.1029/2020JA028505
25 of 30
Journal of Geophysical Research: Space Physics
as an illustrational example. Classical views of the NO production in the nightside thermosphere usually
focus on the following reaction (e.g., Barth etal.2003, 2009; Lin etal.,2018),
2
N O NO O
(R1)
which has a low activation barrier (∼0.27eV) and thus may occur regularly under normal thermosphere
conditions (e.g., with N (2D) participating). However, a major problem of R1 lies in that, even under a
steady-state limit, the resultant NO concentration can only be no more than ∼0.1 (O2) (Hyman etal.,1976).
It is unlikely that the reaction R1 could be capable of producing enough NO and in turn the NO2 continuum
yield to account for the observed STEVE AGC brightness (Gillies etal.,2019; Liang etal.,2019). The other
reaction
2
N O NO N
(R2)
has an activation barrier of ∼3.27eV. To overcome this energy barrier, an electronically excited state of N2 or
O, or a higher-order vibrationally excited state of N2 is required. More specially, for vibrationally excited N2
v′>11 is needed if reacting with ground-state O (3P), or v′>4 if reacting with O (1D), which is known to ex-
ist based upon the 630nm red-line component of STEVE (Gillies etal.,2019; Liang etal.,2019; MacDonald
etal.,2018). Despite the energy barrier, the reaction R2 can potentially yield higher rates of NO production
than R1 (Harding etal.,2020), and was thus at times invoked in attempts to explain the exceptionally high
NO densities in certain observations (e.g., Zipf etal.,1970). In the presence of a strong SAID, N2 molecules
are indeed prone to be excited to higher vibrational levels due to the very high electron temperature (New-
ton etal.,1974), and/or via nonreactive collisions with fast-streaming ions (Harding etal.,2020). Still, the
model is contingent upon the local ambient N2 density. To be capable of producing the realistic STEVE
brightness, a substantial increase of the N2 density in the >150km thermosphere was premised in Harding
etal.(2020), which in their 1D model was assumed to be led by the hydrostatic expansion/upwelling of the
atmosphere under SAID-related heating (e.g., Wang etal.,2012). While such neutral upwelling might in-
deed occur, we suggest that the horizontal wind convergence may serve as one other effective way to help N2
density buildup. Furthermore, as afore-mentioned, the neutral winds may act to transport the vibrationally
excited N2 from the auroral region and pileup at subauroral latitudes. Note that such a transport process
was invoked by Vlasov and Kelley(2003) to explain the formation of the electron trough in the subauroral
region. The vibrationally excited N2 typically has much longer lifetime in the upper thermosphere (∼hours,
see Vlasov and Kelley[2003]) due to the reduced collisional deexcitation with O. When they are transported
into the SAID channel, they may undergo excited-to-excited transitions between different vibrational levels
(Newton etal.,1974), and thus may further populate the higher-level vibrational states of N2. To summarize,
when the above transport processes of excited N2 of auroral origin are considered, the generation of NO via
reaction R2 would be more effective and productive than that evaluated in Harding etal.(2020). At last, we
mention that the neutral winds may also directly help the transport and pileup of auroral precipitation-in-
duced NO at subauroral latitudes (e.g., Solomon etal.,1999).
While our above discussion is mostly built upon a premise that the NO2 AGC contributes to the STEVE
AGC, the essence of our proposal might also be applicable to other possible mechanisms of the chemilumi-
nescence AGC. In short, the presence of SAID, along with its exceptionally high electric field, ion kinetic en-
ergy, and electron temperature, serves as the energy source for STEVE to occur, while a transport and pileup
effect led by the neutral wind variations may prepare a reservoir of relevant neutral constituents and in turn
enhance the production rate and the resultant airglow yield. We thus speculate that STEVE is more readily
to transpire under a coincidental circumstance that the buildup of neutrals occurs around at the same lat-
itude as SAID (though the former may have a somewhat wider latitudinal scale). This may partly explain
the fact that only a subset of SAIDs is found to be accompanied by STEVE, even with conjunctive optical
observations (Archer etal.,2019a; Nishimura etal.,2020a). One example in our event pool is the February
16, 2010 event (event B in Figure10; DMSP data are studied by Nishimura etal.[2020a]). Despite being a
relatively strong substorm (AL∼−570) no equatorward wind transport or divergence is found at subauroral
latitudes. DMSP F18 crossed a SAID channel at ∼08:06UT (though in the southern hemisphere), but in its
conjugate Alaska sector in the northern hemisphere, no clue of STEVE was found in optical data.
LIANG ET AL.
10.1029/2020JA028505
26 of 30
Journal of Geophysical Research: Space Physics
Admittedly, the above proposal remains tentative and somehow hypothetical at the current stage. We em-
phasize again that STEVE is a fairly new phenomenon, and its generation mechanisms remain poorly un-
derstood to date. This paper presents an initial study to address the possible role of neutral dynamics in
STEVE production. To carry on this study, we are to collect more STEVE events with neutral measurements,
from both ground-based instruments and in situ satellites, in the future. We are also working on a global
MIT simulation to model the neutral density/temperature/wind variations during substorm intervals and in
the presence of strong SAPS/SAID. Through these studies, we look forward to gaining a better understand-
ing of the neutral dynamics and its potential contribution to STEVE production in the future.
5. Conclusions
In this study, we make an initial effort to investigate the potential preconditioning role of neutral dynamics
in STEVE production. Using joint SDI and optical ASI observations, we study the pattern and variations
of the neutral wind after the substorm onset but before the STEVE emergence, and compare with that in
nonSTEVE events. Neutral winds enhance in westward and southward directions following substorm au-
roral intensification, and show equatorward propagating trend originating from auroral latitudes. However,
in STEVE events the enhanced equatorward winds appear to feature a stop/reversal at certain subauroral
latitude, and strong wind convergence is developed there. This pattern sustains for ∼15–20min, and then
STEVE arises at about this stop latitude. Such a stop latitude of equatorward winds and wind convergence
are weaker in nonSTEVE events. This is owing partly to that the southward wind intensification from au-
roral latitudes is in general weaker in nonSTEVE events, and partly to that the southward winds are more
liable to penetrate to lower latitudes in nonSTEVE events. The former discrepancy is likely related to the
differences in the underlying substorm strength and/or in the duskward extension of the magnetospher-
ic energy input between STEVE and nonSTEVE substorm events (Nishimura etal.,2020c). We speculate
that a mid-latitude Tn intensification and its resultant southward temperature gradient, which are indeed
observed in STEVE events before the STEVE emergence, might act to impede the southward winds from
auroral latitudes and help form the stop latitude of winds, but such a hunch needs to be furthered examined
in the future, ideally with more definite Tn observations at mid-latitudes.
Our results may shed some implications on the so-far mysterious production mechanism of STEVE. While
there is mounting evidence that the AGC component of STEVE that dominates the overall STEVE bright-
ness (Gillies etal.,2019; Liang etal.,2019), likely stems from certain extremely intensified chemilumines-
cence in the upper thermosphere, many proposals of the underlying AGC mechanisms are baffled by the
requirement of high concentrations of some key neutral continents in subauroral latitudes at STEVE alti-
tudes, to explain the realistic STEVE brightness. We propose that the transport effects led by neutral winds
might partly help solve this difficulty. Enhanced equatorward neutral winds can transport relevant neutrals
species that are excited by auroral precipitation and have crucial importance in the AGC production to sub-
auroral latitudes. These neutrals may pile up due to the wind convergence at the stop latitude of the equa-
torward winds. Such a transport and pileup effect led by the neutral wind variations may prepare a reservoir
of key neutral constituents. When SAID passes through, due to its high ion kinetic energy and electron tem-
perature the SAID would actively interact with the enhanced concentrations of relevant neutrals there (e.g.,
Harding etal.,2020), leading to a dramatic increase of the airglow production and the STEVE occurrence.
Data Availability Statement
SDI data are available at http://sdi_server.gi.alaska.edu/sdiweb. PFISR data are available from http://isr.sri.
com/madrigal. We thank the help of Dr. Donald Hampton in processing the DASC data at GAKO and MSP
data at PKR (available at http://optics.gi.alaska.edu/optics) for some events. We also gratefully acknowledge
the SuperMAG collaborators. We thank for enlightening discussions with Dr. Brian Harding, Dr. J.-P. St.
Maurice, and Dr. Yongliang Zhang.
References
Anderson, C., Conde, M., & McHarg, M. G. (2012a). Neutral thermospheric dynamics observed with two scanning Doppler imagers: 1.
Monostatic and bistatic winds. Journal of Geophysical Research: Space Physics, 117(A3), A03304. https://doi.org/10.1029/2011JA017041
LIANG ET AL.
10.1029/2020JA028505
27 of 30
Acknowledgments
This work is supported by the Canadian
Space Agency. Dr. Ying Zou acknowl-
edges support from the NSF grant
AGS-1502436. The THEMIS mission
is supported by NASA, and data are
available at https://themis.ssl.berkeley.
edu/index.shtml.
Journal of Geophysical Research: Space Physics
Anderson, C., Conde, M., & McHarg, M. G. (2012b). Neutral thermospheric dynamics observed with two scanning Doppler imagers: 2.
Vertical winds. Journal of Geophysical Research: Space Physics, 117(A3), A03305. https://doi.org/10.1029/2011JA017157
Anderson, P. C., & Landry, R. G. (2017). SAPS and SAID: Differences and implications on modeling. SA41C-01, paper presented at 2017 Fall
Meeting, AGU, New Orleans, 11–15 December.
Archer, W. E., Gallardo-Lacourt, B., Perry, G. W., St.-Maurice, J.-P., Buchert, S. C., & Donovan, E. F. (2019a). Steve: The optical signature of
intense subauroral ion drifts. Geophysical Research Letters, 46(12), 6279–6286. https://doi.org/10.1029/2019GL082687
Archer, W. E., & Knudsen, D. J. (2018). Distinguishing subauroral ion drifts from Birkeland current boundary flows. Journal of Geophysical
Research: Space Physics, 123(1)819–826. https://doi.org/10.1002/2017JA024577
Archer, W. E., St.-Maurice, J.-P., Gallardo-Lacourt, B., Perry, G. W., Cully, C. M., Donovan, E., et al. (2019b). The vertical distribu-
tion of the optical emissions of a Steve and Picket Fence event. Geophysical Research Letters, 46(19), 10719–10725. https://doi.
org/10.1029/2019GL084473
Baker, K. B., & Wing, S. (1989), A new magnetic coordinate system for conjugate studies at high latitudes. Journal of Geophysical Research,
94(A7), 9139–9143. https://doi.org/10.1029/JA094iA07p09139
Barbier, D., Dufay, J., & Williams, D. (1951). Recherches sur l'émission de la raie verte de la lumière du ciel nocturne. Annual Review of
Astrophysics, 2(28), 399–437.
Barth, C. A., Lu, G, & Roble, R. G. (2009). Joule heating and nitric oxide in the thermosphere. Journal of Geophysical Research, 114(A5),
A05301. https://doi.org/10.1029/2008JA013765
Barth, C. A., Mankoff, K. D., Bailey, S. M., & Solomon, S. C. (2003). Global observations of nitric oxide in the thermosphere. Journal of
Geophysical Research, 108(A1), 1027. https://doi.org/10.1029/2002JA009458
Cai, L., Oyama, S.-I., Aikio, A., Vanhamäki, H., & Virtanen, I. (2019). Fabry-Perot interferometer observations of thermospheric hori-
zontal winds during magnetospheric substorms. Journal of Geophysical Research: Space Physics, 124(5), 3709–3728. https://doi.
org/10.1029/2018JA026241
Campbell, L., Cartwright, D. C., & Brunger, M. J. (2007). Role of excited N2 in the production of nitric oxide. Journal of Geophysical Re-
search, 112(A8), A08303. https://doi.org/10.1029/2007JA012337
Campbell, L., Cartwright, D. C., Brunger, M. J., & Teubner, P. J. O. (2006). Role of electronic excited N2 in vibrational excitation of the N2
ground state at high latitudes. Journal of Geophysical Research, 111(A9), A09317. https://doi.org/10.1029/2005JA011292
Chu, X., Malaspina, D., Gallardo-Lacourt, B., Liang, J., Andersson, L., Ma, Q., etal. (2019). Identifying STEVE's magnetospheric driver
using conjugate observations in the magnetosphere and on the ground. Geophysical Research Letters, 46(22), 12665–12674. https://doi.
org/10.1029/2019GL082789
Conde, M., & Smith, R. W. (1995). Mapping thermospheric winds in the auroral zone. Geophysical Research Letters, 22(22), 3019–3022.
https://doi.org/10.1029/95GL02437
Conde, M., & Smith, R. W. (1998). Spatial structure in the thermospheric horizontal wind above Poker Flat, Alaska, during solar minimum.
Journal of Geophysical Research, 103(A5), 9471–9949. https://doi.org/10.1029/97JA03331
Dhadly, M., Emmert, J., Drob, D., Conde, M., Doornbos, E., Shepherd, G., etal. (2017). Seasonal dependence of northern high-latitude
upper thermospheric winds: A quiet time climatological study based on ground-based and space-based measurements. Journal of Geo-
physical Research: Space Physics, 122(2), 2619–2644. https://doi.org/10.1002/2016JA023688
Dhadly, M. S., Emmert, J. T., Drob, D. P., Conde, M. G., Doornbos, E., Shepherd, G. G., etal. (2018). Seasonal dependence of geomagnetic
active-time northern high-latitude upper thermospheric winds. Journal of Geophysical Research: Space Physics, 123(1), 739–754. https://
doi.org/10.1002/2017JA024715
Dhadly, M. S., Meriwether, J., Conde, M., & Hampton, D. (2015). First ever cross comparison of thermospheric wind measured by nar-
row- and wide-field optical Doppler spectroscopy. Journal of Geophysical Research: Space Physics, 120(11), 9683–9705. https://doi.
org/10.1002/2015JA021316
Donovan, E. F., Mende, S., Jackel, B., Frey, H., Syrjäsuo, M., Voronkov, I., etal. (2006), The THEMIS all-sky imaging array—System de-
sign and initial results from the prototype imager, Journal of Atmospheric and Solar-Terrestrial Physics, 68(13), 1472–1487. https://doi.
org/10.1016/j.jastp.2005.03.027
Emmert, J. T., Faivre, M. L., Hernandez, G., Jarvis, M. J., Meriwether, J. W., Niciejewski, R. J., et al. (2006). Climatologies of night-
time upper thermospheric winds measured by ground-based Fabry-Perot interferometers during geomagnetically quiet conditions:
1. Local time, latitudinal, seasonal, and solar cycle dependence, Journal of Geophysical Research, 111(A12), A12302. https://doi.
org/10.1029/2006JA011948
Gadsden, M., & Marovich, E. (1973). The nightside continuum. Journal of Atmospheric and Terrestrial Physics, 35(9), 1601–1614. https://
doi.org/10.1016/0021-9169(73)90179-7
Gallardo-Lacourt, B., Liang, J., Nishimura, Y., & Donovan, E. (2018a). On the origin of STEVE: Particle precipitation or ionospheric sky-
glow? Geophysical Research Letters, 45(16), 7968–7973. https://doi.org/10.1029/2018GL078509
Gallardo-Lacourt, B., Nishimura, Y., Donovan, E., Gillies, D. M., Perry, G. W., Archer, W. E., etal. (2018b). A statistical analysis of STEVE.
Journal of Geophysical Research: Space Physics, 123(11), 9893–9905. https://doi.org/10.1029/2018JA025368
Gallardo-Lacourt, B., Nishimura, Y., Lyons, L. R., Mishin, E. V., Ruohoniemi, J. M., Donovan, E. F., etal. (2017). Influence of auroral
streamers on rapid evolution of ionospheric SAPS flows. Journal of Geophysical Research: Space Physics, 122(12), 12406–12420. https://
doi.org/10.1002/2017JA024198
Gillies, D. M., Donovan, E., Hampton, D., Liang, J., Connors, M., Nishimura, Y., etal. (2019). First observations from the TREx Spectro-
graph: The optical spectrum of STEVE and the picket fence phenomena. Geophysical Research Letters, 46(13), 7207–7213. https://doi.
org/10.1029/2019GL083272
Gjerloev, J. W. (2012). The SuperMAG data processing technique. Journal of Geophysical Research: Space Physics, 117(A9), A09213. https://
doi.org/10.1029/2012JA017683
Harding, B. J., Mende, S. B., Triplett, C. C., & Wu, Y.-J. J. (2020). A mechanism for the STEVE continuum emission. Geophysical Research
Letters, 47(7), e2020GL087102. https://doi.org/10.1029/2020GL087102
Hedin, J., Rapp, M., Khaplanov, M., Stegman, J., & Witt, G. (2012). Observations of NO in the upper mesosphere and lower thermosphere
during ECOMA 2010. Annals of Geophysics, 30(11), 1611–1621. https://doi.org/10.5194/angeo-30-1611-2012
Heinselman, C. J., & Nicolls, M. J. (2008). A Bayesian approach to electric field and E-region neutral wind estimation with the Poker Flat
Advanced Modular Incoherent Scatter Radar. Radio Science, 43(5), RS5013. https://doi.org/10.1029/2007RS003805
Hunnekuhl, M., & MacDonald, E. (2019). A citizen science based data package for STEVE phenomenon related subaurora aurora or
auroral-like luminous ionospheric structures. 16th European Space Weather Week. Retrieved from https://register-as.oma.be/esww16/
contributions/public/S2-P1/S2-P1-02-HunnekuhlMichael/POSTER1_ESWW.pdf
LIANG ET AL.
10.1029/2020JA028505
28 of 30
Journal of Geophysical Research: Space Physics
Hunnekuhl, M., & MacDonald, E. (2020). Early ground-based work by auroral pioneer Carl Størmer on the high-altitude detached subau-
roral arcs now known as “STEVE”. Space Weather, 18(3), e2019SW002384. https://doi.org/10.1029/2019SW002384
Hunnekuhl, M., MacDonald, E., Swanson, B., Theusner, M., Stone, J., Chernenkoff, A., etal. (2019). STEVE phenomenon related subau-
roral aurora or aurora-like luminous ionospheric structures—Relevant structures, characteristics and correlations with geomagnetic
storms derived from a citizen science based data package. 16th European Space Weather Week. Retrieved from https://register-as.oma.
be/esww16/contributions/public/S14-P1/S14-P1-05-HunnekuhlMichael/POSTER2_ESWW.pdf
Hyman, E., Strickland, D. J., Julienne, P. S., & Strobel, D. F. (1976), Auroral NO concentrations? Journal of Geophysical Research, 81(25),
4765–4769.
Kwak, Y. S., & Richmond, A. D. (2014). Dependence of the high-latitude lower thermospheric wind vertical vorticity and horizon-
tal divergence on the interplanetary magnetic field. Journal of Geophysical Research: Space Physics, 119(2), 1356–1368. https://doi.
org/10.1002/2013JA019589
Liang, J., Donovan, E., Connors, M., Gillies, D., St-Maurice, J. P., Jackel, B., etal. (2019). Optical spectra and emission altitudes of dou-
ble-layer STEVE: A case study. Geophysical Research Letters, 46(23), 13630–13639. https://doi.org/10.1029/2019GL085639
Lin, C. Y., Deng, Y., Venkataramani, K., Yonker, J., & Bailey, S. M. (2018). Comparison of the thermospheric nitric oxide emission observa-
tions and the GITM simulations: Sensitivity to solar and geomagnetic activities. Journal of Geophysical Research: Space Physics, 123(12),
10239–10253. https://doi.org/10.1029/2018JA025310
Lu, G., Mlynczak, M. G., Hunt, L. A., Woods, T. N., & Roble, R. G. (2010), On the relationship of Joule heating and nitric oxide radiative
cooling in the thermosphere. Journal of Geophysical Research, 115(A5), A05306. https://doi.org/10.1029/2009JA014662
Lyons, L. R., Nishimura, Y., Gallardo-Lacourt, B., Nicolls, M. J., Chen, S., Hampton, D. L., etal. (2015). Azimuthal flow bursts in the inner
plasma sheet and possible connection with SAPS and plasma sheet earthward flow bursts. Journal of Geophysical Research: Space Phys-
ics, 120(6), 5009–5021. https://doi.org/10.1002/2015JA021023
MacDonald, E. A., Donovan, E., Nishimura, Y., Case, N., Gillies, D. M., Gallardo-Lacourt, B., etal. (2018). New science in plain sight:
Citizen scientists lead to the discovery of optical structure in the upper atmosphere. Science Advances, 4(3), eaaq0030. https://doi.
org/10.1126/sciadv.aaq0030
Mishin, E., Nishimura, Y., & Foster, J. (2017). SAPS/SAID revisited: A causal relation to the substorm current wedge. Journal of Geophysical
Research: Space Physics, 122(8), 8516–8535, https://doi.org/10.1002/2017JA024263
Moffett, R. J., Ennis, A. E., Bailey, G. J., Heelis, R. A., & Brace, L. H. (1998). Electron temperatures during rapid subauroral ion drift events.
Annals of Geophysics, 16(4), 450–459.
Newton, G. P., Walker, J. C. G., & Meijer, P. H. E. (1974). Vibrationally excited nitrogen in stable auroral red arcs and its effect on ionospher-
ic recombination. Journal of Geophysical Research, 79(25), 3807–3818.
Nishimura, Y., Donovan, E. F., Angelopoulos, V., & Nishitani, N. (2020a). Dynamics of auroral precipitation boundaries associated with
STEVE and SAID. Journal of Geophysical Research: Space Physics, 125(8), e2020JA028067. https://doi.org/10.1029/2020JA028067
Nishimura, Y., Gallardo-Lacourt, B., Zou, Y., Mishin, E., Knudsen, D. J., Donovan, E. F., etal (2019). Magnetospheric signatures of STEVE:
Implication for the magnetospheric energy source and inter-hemispheric conjugacy. Geophysical Research Letters, 46(11), 5637–5644.
https://doi.org/10.1029/2019GL082460
Nishimura, Y., Lyons, L. R., Gabrielse, C., Sivadas, N., Donovan, E. F., Varney, R. H., etal. (2020b). Extreme magnetosphere-ionosphere-ther-
mosphere responses to the 5 April 2010 supersubstorm. Journal of Geophysical Research: Space Physics, 125(4), e2019JA027654. https://
doi.org/10.1029/2019JA027654
Nishimura, Y., Wygant, J., Ono, T., Iizima, M., Kumamoto, A., Brautigam, D., & Friedel, R. (2008), SAPS measurements around the mag-
netic equator by CRRES, Geophysical Research Letters, 35(10), 1–5. https://doi.org/10.1029/2008GL033970
Nishimura, Y., Yang, J., Weygand, J. M., Wang, W., Kosar, B., Donovan, E. F., etal. (2020c). Magnetospheric conditions for STEVE and SAID:
Particle injection, substorm surge and field-aligned currents. Journal of Geophysical Research: Space Physics, 125(8), e2020JA027782.
https://doi.org/10.1029/2020JA027782
Ritter, P., Lühr, H., & Doornbos, E. (2010. Substorm-related thermospheric density and wind disturbances derived from CHAMP observa-
tions. Annales Geophysicae, 28(6), 1207–1220. https://doi.org/10.5194/angeo-28-1207-2010
Solomon, S., Barth, C. A., & Bailey, S. M. (1999). Auroral production of nitric oxide measured by the SNOE satellite. Geophysical Research
Letters, 26(9), 1259–1262.
Spiro, R. W., Heelis, R. A., & Hanson, W. B. (1979). Rapid subauroral ion drifts observed by Atmospheric Explorer C. Geophysical Research
Letters, 6(8), 657–660.
Sternberg, J. R., & Ingham, M. F. (1972). Observations of the airglow continuum. Monthly Notices of the Royal Astronomical Society, 159(1),
1–20.
Thayer, J. P., & Killeen, T. L. (1991). Vorticity and divergence in the high-latitude upper thermosphere. Geophysical Research Letters, 8(4),
701–704. https://doi.org/10.1029/91GL00131
Wang, W., Talaat, E. R., Burns, A. G., Emery, B., Hsieh, S., Lei, J., & Xu, J. (2012). Thermosphere and ionosphere response to subau-
roral polarization streams (SAPS): Model simulations. Journal of Geophysical Research: Space Physics, 117(A7), A07301. https://doi.
org/10.1029/2012JA017656
Whiter, D. K., Gustavsson, B., Partamies, N., & Sangalli, L. (2013). A new automatic method for estimating the peakauroral emission
height from all-sky camera images. Geoscientific Instrumentation, Methods and Data Systems 2(1), 131–144. https://doi.org/10.5194/
gi-2-131-2013
Vlasov, M. N., & Kelley, M. C. (2003). Modeling of the electron density depletion in the storm-time trough on April 20, 1985, Journal of
Atmospheric and Solar-Terrestrial Physics, 65(2), 211–217. https://doi.org/10.1016/S1364-6826(02)00292-4
Xu, H., Shiokawa, K., Oyama, S. I., & Otsuka, Y. (2019). Thermospheric wind variations observed by a Fabry-Perot interferometer at
Tromsø, Norway, at substorm onsets. Earth, Planets and Space, 71(1), 93. https://doi.org/10.1186/s40623-019-1072-0
Young, R. A. (1969), Chemiluminescent reactions in the airglow, Canadian Journal of Physics, 47(10), 1927–1937.
Zhang, Y., Paxton, L. J., Morrison, D., Marsh, D., & Kil, H. (2014). Storm-time behaviors of O/N2 and NO variations. Journal of Atmospheric
and Solar-Terrestrial Physics, 114, 42–49. https://doi.org/10.1016/j.jastp.2014.04.003
Zipf, E. C., Borst, W. L., & Donahue, T. M. (1970). A mass spectrometer observation of NO in an auroral arc. Journal of Geophysical Re-
search, 75(31), 6371–6376. https://doi.org/10.1029/JA075i031p06371
Zou, S., Lyons, L. R., & Nishimura, Y. (2012). Mutual evolution of aurora and ionospheric electrodynamic features near the Harang reversal
during substorms. Geophysical Monograph Series, 197, 159–169. https://doi.org/10.1029/2011GM001163
LIANG ET AL.
10.1029/2020JA028505
29 of 30
Journal of Geophysical Research: Space Physics
Zou, S., Lyons, L. R., Wang, C.-P., Boudouridis, A., Ruohoniemi, J. M., Anderson, P. C., etal. (2009). On the coupling between the Harang
reversal evolution and substorm dynamics: A synthesis of SuperDARN, DMSP, and IMAGE observations. Journal of Geophysical Re-
search, 114(A1), A01205. https://doi.org/10.1029/2008JA013449
Zou, Y., Lyons, L., Conde, M., Varney, R., Angelopoulos, V., & Mende, S. (2020). Effects of Substorms on High-Latitude Upper Thermo-
spheric Winds. Journal of Geophysical Research: Space Physics, 126(1), e2020JA028193. https://doi.org/10.1029/2020JA028193
Zou, Y., Nishimura, Y., Lyons, L., Conde, M., Varney, R., Angelopoulos, V., & Mende, S. (2018). Mesoscale F region neutral winds associated
with quasi-steady and transient nightside auroral forms. Journal of Geophysical Research: Space Physics, 123(9), 7968–7984. https://doi.
org/10.1029/2018JA025457
References From the Supporting Information
Donovan, E., Spanswick, E., Liang, J., Grant, J., Jackel, B., & Greffen, M. (2012). Magnetospheric dynamics and the proton aurora. In A.
Keiling, E. Donovan, F. Bagenal, T. Karlsson (Eds.), Auroral Phenomenology and Magnetospheric Processes: Earth and Other Planets (pp.
365–378). Washington, DC: AGU.
Immel, T. J., Mende, S. B., Frey, H. U., Peticolas, L. M., Carlson, C. W., Gerard, J. C., etal. (2002). Precipitation of auroral protons in de-
tached arcs, Geophysical Research Letters, 29(11), 14-1–14-4. https://doi.org/10.1029/2001GL013847
Lyons, L. R., Nishimura, Y., Gallardo-Lacourt, B., Nicolls, M. J., Chen, S., Hampton, D. L., etal. (2015). Azimuthal flow bursts in the inner
plasma sheet and possible connection with SAPS and plasma sheet earthward flow bursts. Journal of Geophysical Research: Space Phys-
ics, 120(6), 5009–5021. https://doi.org/10.1002/2015JA021023
Nishimura, Y., Bortnik, J., Li, W., Lyons, L. R., Donovan, E. F., Angelopoulos, V., & Mende, S. B. (2014). Evolution of nightside subauroral
proton aurora caused by transient plasma sheet flows. Journal of Geophysical Research: Space Physics, 119(7), 5295–5304. https://doi.
org/10.1002/2014JA020029
Wallis, D. D., Burrows, J. R., Moshupi, M. C., Anger, C. D., & Murphree, J. S. (1979). Observations of particles precipitating into de-
tached arcs and patches equatorward of the auroral oval, Journal of Geophysical Research, 84(A4), 1347–1360. https://doi.org/10.1029/
JA084iA04p01347
Zhang, Y., Paxton, L. J., Morrison, D., Wolven, B., Kil, H., & Wing, S. (2005), Nightside detached auroras due to precipitating protons/ions
during intense magnetic storms, Journal of Geophysical Research, 110(A2), A02206. https://doi.org/10.1029/2004JA010498
LIANG ET AL.
10.1029/2020JA028505
30 of 30
Content uploaded by Jun Liang
Author content
All content in this area was uploaded by Jun Liang on Mar 03, 2021
Content may be subject to copyright.
