Auroral Image Data Research Articles

Classification of Ground‐Based Auroral Images by Learning Deep Tensor Feature Representation on Riemannian Manifold

AbstractAutomatically classifying a huge amount of ground‐based auroral images is essential to facilitate aurora morphology statistical research and aid in comprehending the magnetospheric dynamics. However, facing the challenge of insufficient labeled images, deep learning methods perform suboptimally on small auroral image data sets, and traditional machine learning methods based on handcrafted features heavily rely on expert knowledge. In this paper, we propose a novel method that leverages the merits of both traditional machine learning and deep learning methods by extracting deep second‐order tensor features to train a support vector machine (SVM). To improve compactness and discriminative ability of the features, we comply the intrinsic data geometry on Riemannian manifold to employ dimensionality reduction and map the dimensionality‐reduced features from Riemannian space to Euclidean space for the SVM classifier. Experimental results on small aurora data sets conclusively demonstrate the effectiveness of our method, exhibiting competitive performance with recent aurora classification methods.

Auroral breakup detection in all-sky images by unsupervised learning

Abstract. Due to a large number of automatic auroral camera systems on the ground, image data analysis requires more efficiency than what human expert visual inspection can provide. Furthermore, there is no solid consensus on how many different types or shapes exist in auroral displays. We report the first attempt to classify auroral morphological forms by an unsupervised learning method on an image set that contains both nightside and dayside aurora. We used 6 months of full-colour auroral all-sky images captured at a high-Arctic observatory on Svalbard, Norway, in 2019–2020. The selection of images containing aurora was performed manually. These images were then input into a convolutional neural network called SimCLR for feature extraction. The clustered and fused features resulted in 37 auroral morphological clusters. In the clustering of auroral image data with two different time resolutions, we found that the occurrence of 8 clusters strongly increased when the image cadence was high (24 s), while the occurrence of 14 clusters experienced little or no change with changes in input image cadence. We therefore investigated the temporal evolution of a group of eight “active aurora” clusters. Time periods for which this active aurora persisted for longer than two consecutive images with a maximum cadence of 6 min coincided with ground-magnetic deflections, and their occurrence was found to maximize around magnetic midnight. The active aurora onsets typically included vortical auroral structures and equivalent current patterns typical for substorms. Our findings therefore suggest that our unsupervised image clustering method can be used to detect auroral breakups in ground-based image datasets with a temporal accuracy determined by the image cadence.

SAPS‐Associated Explosive Brightening on the Duskside: A New Type of Onset‐Like Disturbance

AbstractQuasiperiodic energetic particle injections have been observed at geosynchronous orbit on the duskside during a steady magnetospheric convection event. We examine high‐resolution auroral imager data and ground magnetometer data associated with the first of these injections and conclude that it was not associated with classical substorm signatures. It is proposed that these injections are caused by the explosive nonlinear growth of a shear flow‐ballooning instability in the region where subauroral polarization streams (SAPS) also occur. It is suggested that interchange will occur preferentially in the low‐conductivity SAPS region since the magnetic Richardson number is lowest there and the “line‐tying” effect will also be least stabilizing there. We propose that the observed particle injection signatures and auroral morphology constitute a new type of SAPS‐associated explosive “onset‐like” disturbance that can occur during intervals of strong convection.

Tracking patchy pulsating aurora through all-sky images

Abstract. Pulsating aurora is frequently observed in the evening and morning sector auroral oval. While the precipitating electrons span a wide range of energies, there is increasing evidence that the shape of pulsating auroral patches is controlled by structures in near-equatorial cold plasma; these patches appear to move with convection, for example. Given the tremendous and rapidly increasing amount of auroral image data from which the velocity of these patches can be inferred, it is timely to develop and implement techniques for the automatic identification of pulsating auroral patch events in these data and for the automatic determination of the velocity of individual patches from that data. As a first step towards this, we have implemented an automatic technique for determining patch velocities from sequences of images from the Time History of Events and Macroscale Interactions during Substorms (THEMIS) all-sky imager (ASI) and applied it to many pulsating aurora events. Here we demonstrate the use of this technique and present the initial results, including a comparison between ewograms (east–west keograms) and time series of patch position as determined by the algorithm. We discuss the implications of this technique for remote sensing convection in the inner magnetosphere.

Auroral spectral estimation with wide-band color mosaic CCDs

Abstract. Optical aurora can be structured over a wide range of spatial and temporal scales with spectral features that depend on the energy of precipitating particles. Scientific studies typically combine data from multiple instruments that are individually optimized for spatial, spectral, or temporal resolution. One recent addition combines all-sky optics with color mosaic CCD (charge-coupled device) detectors that use a matrix of different wide-band micro-filters to produce an image with several (often three) color channels. These devices provide sequences of two dimensional multispectral luminosity with simultaneous exposure of all color channels allowing interchannel comparison even during periods with rapidly varying aurora. At present color auroral image data are primarily used for qualitative analysis. In this study a quantitative approach based on Backus–Gilbert linear inversion was used to better understand the effective spectral resolution of existing and proposed instruments. Two spectrally calibrated commercial detectors (Sony ICX285AQ and ICX429AKL) with very different color mosaics (RGB (red, green, blue) vs. CYGM (cyan, yellow, green, magenta)) were found to have very similar spectral resolution: three channels with FWHM (full-width half-maximum) ≈100 nm; a NIR (near infrared) blocking filter is important for stabilizing inversion of both three-channel configurations. Operating the ICX429AKL in a noninterlaced mode would improve spectral resolution and provide an additional near infrared channel. Transformations from arbitrary device channels to RGB are easily obtained through inversion. Simultaneous imaging of multiple auroral emissions may be achieved using a single-color camera with a triple-pass filter. Combinations of multiple cameras with simple filters should provide ~50 nm resolution across most of the visible spectrum. Performance of other instrument designs could be explored and compared using the same quantitative framework.

A state‐of‐the‐art picture of substorm‐associated evolution of the near‐Earth magnetotail obtained from superposed epoch analysis

We have obtained a state‐of‐the‐art picture of substorm‐associated evolution of the near‐Earth magnetotail and the inner magnetosphere for understanding the substorm triggering mechanism. We performed superposed epoch analysis of Geotail, Polar, and GOES data with 2‐min resolution, utilizing a total of 3787 substorms for each of which auroral breakup was determined from Polar UVI or IMAGE FUV auroral imager data. The decrease of the north‐south magnetic field associated with plasmoids and the initial total pressure decrease suggest that the magnetic reconnection first occurs in the premidnight tail, on average, at X ∼ −16 to −20 RE at least 2 min before auroral onset. The magnetic reconnection site is located near the tailward edge of a region of considerably taillike magnetic field lines and intense cross‐tail current, which extends from X ∼ −5 to −20 RE in the premidnight sector. Then the plasmoid substantially evolves tailward of X ∼ −20 RE immediately after onset. Almost simultaneously with the magnetic reconnection, the dipolarization begins first at X ∼ −7 to −10 RE 2 min before onset. The dipolarization region then expands tailward as well as in the dawn‐dusk directions and earthward. We find that the total pressure generally enhances in association with the dipolarization, with the contribution of high‐energy particles. Also, energy release is more significant between the regions of the magnetic reconnection and the initial dipolarization. The present results will be helpful as a reference guide to developing the overall picture of magnetotail evolution and studying the causal relationship between the magnetic reconnection and the dipolarization as well as detailed mechanisms of each of the two processes on the basis of multispacecraft observations.

Longitudinal dependence of characteristics of low-latitude Pi2 pulsations observed at Kakioka and Hermanus

AbstractWe statistically investigated longitudinal dependence of characteristics of low-latitude Pi2 pulsations to find the longitudinal structure of the plasmaspheric cavity mode. We used the geomagnetic field data from two ground stations, Kakioka (27.2° geomagnetic latitude, 208.5° geomagnetic longitude) and Hermanus (—33.9° geomagnetic latitude, 82.2° geomagnetic longitude), and auroral image data acquired by the ultraviolet imager onboard the Polar satellite for the period of December 4, 1996 to March 3, 1997. Our findings include the following: (1) Pi2 amplitude is the largest around the magnetic local time of the auroral breakup site and decreases away from it; (2) when a nightside Pi2 pulsation has large amplitude, a dayside Pi2 pulsation can be observed with a similar waveform; (3) Pi2 pulsations generally have no clear phase differences (mean phase difference of 3.3°) between Kakioka and Hermanus, except for some events; and (4) the phase difference is independent on AMLT (difference of magnetic local time between a station and the auroral breakup). These observations suggest that the plasmaspheric cavity mode can be excited globally with a very small value of the azimuthal wave number(m≈0).

Influence of ionosphere conductivity on the ring current

Using the Comprehensive Ring Current Model (CRCM), which self‐consistently solves the kinetic equation of ring current protons and the closure of the electric current between the magnetosphere and ionosphere, we have studied how different changes in the ionospheric conductivity affect the strength of the ring current. The conductivity for F10.7 = 250 × 104 Janskys (Jy) (solar maximum condition) results in a ring current that is about 29% stronger than for F10.7 = 70 × 104 Jy (solar minimum condition). The conductivity at equinox results in a ring current that is about 5% stronger than at solstice because the two‐hemisphere height‐integrated conductivities at equinox are higher than at solstice. This would be a new mechanism for explaining the semiannual variation of Dst. Simulation with a realistic auroral conductivity estimated from the Imager for Magnetopause‐to‐Aurora Global Exploration (IMAGE)/Far Ultraviolet Imager (FUV) auroral imager data reveals the fact that auroral brightenings do not significantly change the intensity of the ring current. The overshielding condition is found to be produced when the auroral conductivity decreases abruptly near the Dst minimum, triggering a rapid decay of the ring current. The ring current is shown to be influenced not only by the interplanetary magnetic field and the solar wind but also by solar radiation and morphological features of the auroral electron precipitation as well.

Diurnal auroral occurrence statistics obtained via machine vision

Abstract. Modern ground-based digital auroral All-Sky Imager (ASI) networks capture millions of images annually. Machine vision techniques are widely utilised in the retrieval of images from large data bases. Clearly, they can play an important scientific role in dealing with data from auroral ASI networks, facilitating both efficient searches and statistical studies. Furthermore, the development of automated techniques for identifying specific types of aurora opens up the potential of ASI control software that would change instrument operation in response to evolving geophysical conditions. In this paper, we describe machine vision techniques that we have developed for use on large auroral image data sets. We present the results of application of these techniques to a 350000 image subset of the CANOPUS Gillam ASI in the years 1993–1998. In particular, we obtain occurrence statistics for auroral arcs, patches, and Omega-bands. These results agree with those of previous manual auroral surveys.Key words. Ionosphere (Instruments and techniques) General (new fields)

Interhemispheric asymmetry of the high‐latitude ionospheric convection on 11–12 May 1999

Ionospheric convection over the southern polar cap on 11–12 May 1999 has been studied by using the Syowa East and South HF radar data and the DMSP ion driftmeter data, when the solar wind density was very low and geomagnetic activity was low. The overall convection pattern is consistent with the previous results by Ohtani et al. [2000]. However, the Syowa radars and the DMSP satellites observed very high (>1500 m/s) westward plasma flows at dusk directed from the nightside toward the dayside only in the Southern (dark) Hemisphere. The high‐speed flow was observed continuously across the fields of view of both radars from 1530 UT on 11 May to 0200 UT on 12 May, when the solar wind density was close to minimum. Comparison with the DMSP particle and auroral image data shows that the westward flow regions were located in the middle of the auroral precipitation area. The strong asymmetry of the convection between the two hemispheres indicates the importance of the presence (absence) of solar illumination for the absence (presence) of the strong and localized ionospheric flows.

Development of the multi-spectral auroral camera onboard the index satellite

To investigate the fine-scale auroral structures, high time and spatial resolution imaging observations of optical auroras will be made by a multi-spectral auroral camera (MAC) onboard the INDEX satellite which will be launched by an H2A rocket as a piggyback satellite into a polar orbit at an altitude of ∼700 km. Monochromatic auroral image data at emissions of N2 + first negative band (427.8 nm), OI (557.7 nm), and N 2 first positive band (670 nm) are obtained by MAC with the field-of-view (FOV) of 7.6° using three independent CCD cameras in combination with interference filters. MAC will operate in the nightside auroral region by two operation modes in the following. (1) Simultaneous measurement with particle sensors (ESA/ISA). In this mode, MAC observes an imaging area of ∼80×80 km (at a 100 km altitude) around a magnetic footprint with spatial and time resolutions of ∼1.2 km and 120 msec, respectively. (2) Auroral height distribution measurement. The attitude of INDEX satellite is changed to direct the FOV of MAC on the limb of the Earth. In this mode, MAC observes an imaging area of ∼270×270 km (at a 2000 km distance from the satellite) with spatial and time resolutions of ∼4 km and 1 sec, respectively. In this paper, the science mission, the instrumentation, and observation modes concerning on MAC will be presented.

Midday sub‐auroral patches (MSPs) associated with interplanetary shocks

We report a new category of dayside auroral transients using auroral image data from the Polar spacecraft Ultraviolet Imager. The transients are characterized by a sudden brightening of the auroral patch in the midday sub‐auroral zone, and thus are named midday sub‐auroral patches (MSPs). MSPs were found to occur concurrently with storm sudden commencements in response to the compression of the magnetosphere by interplanetary (IP) shocks. After brightening, the MSP widened and/or moved equatorward. The latitudinal location of an MSP depends on the solar wind dynamic pressure in the shock downstream: a larger dynamic pressure corresponds to a lower latitude. The lifetime of an MSP is short (∼5–6 min). The magnetospheric source region for the MSPs, as determined with the Defense Meteorological Satellite Program particle data from one fortuitous event, is the central plasma sheet (CPS). Sudden compression of the Earth's magnetic field due to the IP shock impact may change the pitch angles of a pre‐existing population of CPS particles, which consequently fall into the loss cone and produce auroral emission.

Polar CEPPAD/IPS energetic neutral atom (ENA) images of a substorm injection

The CEPPAD Imaging Proton Spectrometer on the POLAR spacecraft has proven to perform very well as an Energetic Neutral (ENA) atom imager, despite the fact that it was designed primarily for measuring energetic ions in-situ. ENAs emitted from the ring current can be detected during storm- as well as quiet-time conditions and can be monitored continuously for many hours at a time when Polar is situated in the polar cap. In addition, we are able to routinely detect ‘bursts’ of ENA emissions in response to substorm-associated ion injections. In this paper, we present ENA images of a single such event together with global auroral imager data from the POLAR VIS instrument. LANL geosynchronous energetic particle data, and ground magnetic Pi2 data in order to establish that such bursts are indeed caused by substorm injections.

Thermospheric wind field over Mawson and Davis, Antarctica; simultaneous observations by two Fabry–Perot spectrometers of λ630 nm emission

The thermopheric oxygen λ630 nm emission has been observed using high-resolution Fabry–Perot spectrometers at Mawson (67.6°S, 62.9°E) and Davis (68.6°S, 78.0°E), Antarctica. A new technique, combining the results from the two instruments, is used to derive vector wind fields. The technique is described and applied to five nights of simultaneous cardinal point data obtained in 1997. Solar flux was low during this interval, typically F 10.7=75. Of the five nights two were magnetically disturbed and three were quiet. The observations for the disturbed nights were compared to a TIEGCM model run and reasonable agreement was found in the first half of the night. On one of the disturbed nights a closed evening circulation cell and cross-polar jet could be identified in our data. On none of the nights was a morning circulation cell evident. Auroral imager data were used to locate the auroral oval. For several hours around magnetic midnight the auroral oval produces doldrums in the thermospheric winds that are not described by the model. Auroral doldrums are also seen on the quiet nights which otherwise maintain a flow approximately consistent with a pressure-gradient driver.

Observations of magnetospheric substorms occurring with no apparent solar wind/IMF trigger

An outstanding topic in magnetospheric physics is whether substorms are always externally triggered by disturbances in either the interplanetary magnetic field (IMF) or solar wind, or whether they can also occur solely as the result of an internal magnetospheric instability. Over the past decade, arguments have been made on both sides of this issue. Horwitz [1985] and McPherron et al. [1986] have shown examples of substorm onsets which they claimed were not externally triggered. However, as pointed out by Lyons [1995, 1996], there are several problems associated with these studies that make their results somewhat inconclusive. In particular, in the McPherron et al. study, fluctuations in the By component were not considered as possible triggers. Furthermore, Lyons suggests that the sharp decreases in the AL index during intervals of steady IMF/solar wind are not substorms at all but rather that they are just enhancements of the convection driven DP 2 current system that are often observed to occur during steady magnetospheric convection events. In the present study, we utilize a much more comprehensive data set (consisting of particle data from the Los Alamos energetic particle detectors at geosynchronous orbit, IMP 8 magnetometer and plasma data, Viking UV auroral imager data, midlatitude Pi 2 pulsation data, ground magnetometer data, and ISEE 1 magnetic field and energetic particle data) to show as unambiguously as possible that typical substorms can indeed occur in the absence of an identifiable trigger in the solar wind/IMF.

Four large‐scale field‐aligned current systems in the dayside high‐latitude region

A system of four current sheets of large‐scale field‐aligned currents (FACs) was discovered in the data set of simultaneous Viking and DMSP‐F7 crossings of the dayside high‐latitude region. This paper reports four examples of this system that were observed in the prenoon sector. The flow polarities of FACs are upward, downward, upward, and downward, from equatorward to poleward. The lowest‐latitude upward current is flowing mostly in the CPS precipitation region, often overlapping with the BPS at its poleward edge, and is interpreted as a region 2 current. The pair of downward and upward FACs in the middle of the structure are collocated with structured electron precipitation. The precipitation of high‐energy (>1 keV) electrons is more intense in the lower‐latitude downward current sheet. The highest‐latitude downward flowing current sheet is located in a weak, low‐energy particle precipitation region, suggesting that this current is flowing on open field lines. Simultaneous observations in the postnoon local time sector reveal the standard three‐sheet structure of FACs, sometimes described as region 2, region 1, and mantle (referred to the midday region 0) currents. A high correlation was found between the occurrence of the four FAC sheet structure and negative interplanetary magnetic field (IMF) By. We discuss the FAC structure in terms of three types of convection cells: the merging, viscous, and lobe cells. During strongly negative IMF By, two convection reversals exist in the prenoon sector; one is inside the viscous cell, and the other is between the viscous cell and the lobe cell. This structure of convection flow is supported by the Viking electric field and auroral UV image data. Based on the convection pattern, the four FAC sheet structure is interpreted as the latitudinal overlap of midday and morning FAC systems. We suggest that the four‐current sheet structure is common in a certain prenoon local time sector during strongly negative IMF By.

Simultaneous prenoon and postnoon observations of three field‐aligned current systems from Viking and DMSP‐F7

The spatial structure of dayside large‐scale field‐aligned current (FAC) systems is examined by using Viking and DMSP‐F7 data. We focus on four events in which the satellites simultaneously observed postnoon and prenoon three FAC systems: the region 2, the region 1, and the mantle (referred to as midday region 0) systems, from equatorward to poleward. These events provide the most solid evidence to date that the midday region 0 system is a separate and unique FAC system, and is not an extension of the region 1 system from other local times. The events are examined comprehensively by making use of a multi‐instrument data set, which includes magnetic field, particle flux, electric field, auroral UV image data from the satellites, and the Sondrestrom convection data. The results are summarized as follows: (1) Region 2 currents flow mostly in the CPS precipitation region, often overlapping with the BPS at their poleward edge. (2) The region 1 system is located in the core part of the auroral oval and is confined in a relatively narrow range in latitude which includes the convection reversal. The low‐latitude boundary layer, possibly including the outer part of the plasma sheet, and the external cusp are the major source regions of dayside region 1 currents. (3) Midday region 0 currents flow on open field lines and are collocated with the shear of antisunward convection flows with velocities decreasing poleward. On the basis of these results we support the view that both prenoon and postnoon current systems consist of the three‐sheet structure when the distortion of the convection pattern associated with interplanetary magnetic field (IMF) By is small and both morningside and eveningside convection cells are crescent‐shaped. We also propose that the midday region 0 and a part of the region 1 systems are closely coupled to the same source.

Effects of ionospheric recombination time scale on the auroral signature of substorms

The magnetosphere‐ionosphere coupling model of substorms developed by Kan et al. (1988) is improved by including the realistic recombination time scale in the ionosphere and the quiet‐time convection prior to the enhancement of the magnetospheric convection. One of the new features of the improved model is the appearance of highly localized structures of the discrete aurora along the poleward boundary of the diffuse aurora in the evening‐midnight sector. This feature is consistent with the auroral image data of the Viking satellite. Our results show that a discrete auroral arc suddenly brightens near midnight and starts moving westward when the auroral brightness increases to a certain level, which is similar to the substorm onset scenario. The bright discrete aurora has an initial velocity of around 2 km s−1, but gradually slows down and dims. This feature is identified with the westward traveling surge. Following passage of the surge, several localized auroral arcs remain along the auroral oval, and the most active auroral region shifts from premidnight toward the evening sector.

Snapshots of high‐latitude electrodynamics using Viking and DMSP F7 observations

Simultaneous observations by the Viking and the DMSP F7 satellites have been used as input to a new method to obtain snapshot pictures of the auroral electrodynamics. In particular, an “instantaneous” global equipotential (or convection) pattern is calculated from distributions of field‐aligned current and conductivity which are qualitatively consistent with the Viking auroral imager data and quantitatively consistent with magnetic field and particle data from the two satellites. This convection pattern, which is of the normal two‐cell type, with a weak dusk cell and a strong, elongated crescent‐shaped dawn cell (consistent with positive interplanetary magnetic field By), agrees well with the Viking electric field data. The model and the observed potential profiles agree nicely along the entire Viking orbit except for two intervals above acceleration regions where deviations are to be expected (due to parallel electric fields). These regions are characterized by U‐shaped potential minima, upward field‐aligned currents, upgoing ion beams, and relatively intense auroral kilometric radiation. Thus, the model results are consistent with the Viking observations not only on a global scale but also on the scale of the auroral acceleration regions. The corresponding convection in the magnetosphere is obtained from a simple projection to the equatorial plane of the deduced two‐cell convection pattern. From this the location of the plasmapause is inferred.

A new method to derive “instantaneous” high‐latitude potential distributions from satellite measurements including auroral imager data

Electric field, magnetic field, auroral imager and particle observations obtained on the Viking satellite for one dusk‐dawn crossing through the polar cusp, are used to illustrate a new promising technique to derive “instantaneous” pictures of the high‐latitude electrodynamics. Global electric field distributions are calculated using as input UV‐imager data of the auroral distribution and in situ measurements of the electrodynamical parameters along the Viking orbit, which are combined with measured average distributions of field‐aligned currents (Region 1 and 2) and conductivities representing the diffuse auroral background. The input data (j∥ and ΣP, ΣH) with contributions from both the diffuse auroral background and active regions, and the output data (E) are tested and calibrated to the Viking measurements by point‐by‐point quantitative comparisons along the magnetic footpoint of the satellite. The calculations are done iteratively to refine and improve both the input and the output. The results demonstrate that selfconsistent and realistic solutions of the prevailing “instantaneous” auroral electrodynamics are possible to obtain in this manner.