Aurorae, as a phenomenon, have been studied extensively over the past century, and mesoscale auroral forms like arcs are generally rather well understood. Some open questions remain, though, such as the intricacies of sudden changes in morphology and the drivers behind dynamic auroral processes . Small-scale features, on the other hand, are much less well known and new features are still being found, for example, transient phenomena such as Lumikot .

Auroral emissions are dependent on the atmospheric composition, which varies with altitude. The same wavelengths that are typically observed with aurorae can also be emitted without the presence of particle precipitation. One such example is airglow, which can produce the same 557.7 nm and 630.0 nm emission lines of atomic oxygen as typical aurorae, but in this case it is due to dissociative electron recombination e.g.. Interaction between aurorae and the dynamics of the neutral atmosphere is a complex subject, with features such as the recently discovered dunes potentially being caused by atmospheric wave modulation on diffuse aurorae . Thus, not all emissions similar to aurorae are caused by particle precipitation; Strong Thermal Emission Velocity Enhancement (STEVE) is already a well-known example of aurora-like skyglow which is likely caused by local acceleration processes instead of precipitation . It is sometimes accompanied by green rays known as the picket fence below the purple arc of STEVE . This picket fence is ostensibly related to particle precipitation , although some studies have questioned this connection based on spectral analysis . STEVE itself has been associated with subauroral ion drifts and local electron heating .

In this study, we suggest fragmented aurora-like emissions (FAEs) as being another phenomenon in the same category of aurora-like phenomena for which particle precipitation is unlikely to be the direct cause. The small fragments of excited plasma discussed in the present study seem to differ from other auroral structures in various ways. They exhibit small horizontal scales of only a few kilometres, short lifetimes of generally less than a minute, and a lack of field-aligned emission extent. Generally, the FAEs occur close to auroral features. This is especially true for FAEs of the second type, occurring in wave-like structures, which were observed with an offset to auroral arcs on the same scale as the FAE size. The next section of the present study aims to provide an overview of the observations and instrumentation used to gather data, followed by a more in-depth description of FAE characteristics. Finally, we suggest some potential generation mechanisms and relations to other recently discovered aurora-like phenomena and summarise our conclusions.

All of the analysed FAEs were observed on all-sky camera (ASC) images captured at the Kjell Henriksen Observatory (KHO), which is located on the Breinosa mountain east of Longyearbyen, Svalbard at ∼ 78.15∘ N, 16.04∘ E. The first observation was made on 7 November 2015 between 20:15:58 and 20:17:27 UTC, with four identified FAE candidates over four images (further referred to as event 1). FAEs were next seen again on 7 December 2015 between 18:18:14 and 18:27:36 UTC (20 images; event 2); this time there were a total of 39 candidates. The final observation that is analysed in the present study was made over a much longer period on 18 December 2017 between 07:36:35 and 08:26:48 UTC, consisting of 79 images (event 3) in which 262 candidates were marked. Figure  shows all these marked candidates for event 3 overlain on the first image of the series taken at 07:36:35 UTC. This is done to visualise the distribution across the sky and the general characteristics of the marked candidates; almost all occurred at a later time during event 3. All FAE events were accompanied by aurorae. It should be noted that FAEs have also been sighted at the KHO on at least three other dates, which were more recent and, thus, not included in the present study.

Due to the availability of varied instrumentation on Svalbard, an effort was made to incorporate many different data sources to obtain FAE characteristics. These include the Sony α7S all-sky camera (ASC) the and meridian-scanning photometer (MSP) at the KHO, as well as data from the European Incoherent Scatter Scientific Association (EISCAT) Svalbard Radar (ESR) and high frame rate optical observations with the Auroral Structure and Kinetics (ASK) instrument located at the ESR. The ASC images used in the present study have a size of 2832 × 2832 pixels. The images were taken using an exposure time of 4 s and an ISO of 16 000 at a cadence of 11 to 12 s, with a mean interval length of 11.8 s. This variance is due to variations of the read-out time to the attached computer. The read-out delay between the camera and software is responsible for the slower cadence, compared to the camera exposure time. A simple astrometry calibration was used to find the centre of the ASC images and estimate the pixel size, resulting in a scale of 16.59 pixel per degree close to centre. This is further used to determine the offset of FAEs from zenith, which can then be used to calculate the pixel sizes in kilometres for varying elevation angles, using an equidistant projection and an assumed FAE altitude of 110 km. This assumption was based on FAE signatures in the ESR data.

Spectral information is provided by the MSP, which scans the auroral emissions at 427.8 (N2+), 557.7 and 630.0 nm (both atomic oxygen) with a 1∘ field of view (FOV) from north to south along the local geomagnetic meridian (31∘ west of geodetic north) using a rotating mirror. Measurements have a time resolution of 8 s (16 s for events 1 and 2), consisting of 4 s (8 s) for a full 360∘ scan plus another 4 s (8 s) for a background scan. Thus, scanning across the sky takes 2 s (4 s). The background measurement is achieved by tilting a narrow bandpass (∼ 0.5 nm) interference filter for each channel .

High temporal resolution optical observations from ASK are used to further study the movement and emission properties of the FAEs. The ASK instrument consists of three channels with individual bandpass filters for selected auroral wavelengths and lenses to adjust FOV . This allows for simultaneous observations of different auroral emissions in a narrow FOV, which can be used to study the energy and flux of the precipitating electrons that produce the aurora . The temporal resolution is 20–32 Hz, and for resolutions above 5 Hz, the available 512 pixels for each camera are binned into a 256×256 pixel image . ASK points towards the magnetic zenith and shares part of its observation region with the ESR and the MSP, which leads to a finding of a FAE signature in the ESR data after observing its passing across the FOV of ASK. The ASK FOV is 6.2∘, and in this study, we use observations of N2 (673.0 nm, first positive band system) and atomic oxygen (777.4 nm) emissions.

Solar wind data from the Advanced Composition Explorer (ACE) and Deep Space Climate Observatory (DSCOVR) satellites at the L1 Lagrangian point can provide insight into the background conditions during the observed events. For the periods preceding the two larger events (2 and 3), the ACE and DSCOVR data show average speeds of 620–640 km s-1, which is above the threshold value for high-speed streams . The Bz component of the interplanetary magnetic field (IMF) is negative, and the IMF By is positive for the relevant periods preceding the FAE occurrences. This indicates an efficient energy transfer into the magnetosphere–ionosphere system. The ACE data for event 1 show average solar wind speeds of ∼540 km s-1, negative IMF Bz, both of which resemble the other two events to some degree, but negative IMF By. The Kp indices for the time periods of events 1–3 are 3+, 4-, and 4+, indicating moderate geomagnetic activity. Visually inspected convection maps from Super Dual Auroral Radar Network (SuperDARN) radars suggest an ionospheric plasma flow primarily in the northwestern or southwestern direction. For all our event times, Svalbard was located in the evening cell of the convection and close to the flow reversal.

FAEs can be categorised into two distinct categories, with the first being individually occurring FAEs. These occur seemingly randomly across the sky, sometimes with a significant offset to the closest auroral arc. The second type are periodic structures with regular spacing between FAEs, which appear close to and generally northwards of auroral arcs. The category 2 FAE group shown in Fig.  is a typical example.

Fragmented aurora-like emissions (FAEs) have been analysed and classified in the present study, with results suggesting that they are a new type of aurora-like feature. Comparing FAEs with ostensibly similar auroral phenomena shows some key differences. For example, the term enhanced aurora EA; see describes an enhanced emission in a thin layer, typically along a rayed auroral structure. Although it also designates a localised emission intensity enhancement occurring alongside aurorae, EA differs in various characteristics. EA occurs as layers with limited vertical extent but has longitudinal and latitudinal extents of at least 250 and 300 km, respectively . FAEs are much smaller, with minor and major axes sizes of < 10 km. While EA manifests as intensity enhancements along the rays of a bigger auroral feature, FAEs were clearly dislocated from the field lines of the adjacent rayed structures. FAEs also lack the blue emission enhancement visible in EA. Furthermore, EA has been observed as quasi-stable structures lasting for minutes, while most analysed FAEs had lifetimes of less than a minute. Overall, this suggests that these are two different phenomena.

When comparing FAEs with pulsating patches, two major distinctions between the two phenomena are size and lifetime of the individual features. Pulsating patches occur within diffuse aurorae, whereas the analysed FAEs are seen alongside discrete arcs. FAEs are much smaller than pulsating patches, which are also typically very stable, while showing quasi-periodic fluctuations in their emission intensity e.g.. In contrast, FAEs are short-lived and do not show any emission intensity fluctuations, apart from appearing and fading away. The available ASK video observations of FAEs show their much higher dynamic motion and smaller size, compared to pulsating patches. Together, these differences lead us to conclude that FAEs are a distinctly different phenomenon.

As the FAEs were found by a manual inspection of images, there is some bias in terms of which features were selected and how they were classified. The data set could contain other auroral small-scale forms or diffuse patches, which is the reason for the classification into four confidence groups. As the general properties of candidates between high- and low-confidence groups agree well, we are confident that most selected features are indeed FAEs. Generally, FAEs can be distinguished from other auroral forms by their lack of field-aligned emission extent, as suggested by the off-zenith parts of the ASC images and field-aligned ionisation measured by the ESR, small sizes and short lifetimes. A FAE signature is visible in the ESR data as locally enhanced electron temperatures around 113 km. Determining a definite FAE altitude requires triangulation, which was not possible for the analysed ASC images or other means of consistently identifying FAE signatures in measurements over an altitude range, such as multiple signatures in EISCAT data.

recently described green streaks below STEVE, which show various similarities to FAEs. Their triangulation positions the streaks at an altitude of 100–110 km, which is also the region within which we suggest FAEs occur. They propose superthermal electrons resulting from the extreme electric fields during STEVE as a local generation mechanism, similar to our hypothesis. It will be interesting to see if these two phenomena are indeed related on a fundamental level or just bear superficial resemblance. suggest STEVE as another locally generated skyglow without any associated particle precipitation. The phenomenon is far from well understood and occurs on much larger scales than FAEs but indicates that ionospheric processes can indeed cause emission without particle precipitation being present. We propose that FAEs fall within the same category, even though many of their properties, such as size and lifetime, differ majorly. The underlying processes heating the plasma are unlikely to be the same, but on a fundamental level, both emissions seem to be related to thermal ionospheric processes rather than particle precipitation.

The present study aims to present the basic characteristics of FAEs and categorise them based on the three analysed events. Nonetheless, the available data enable us to hypothesise about their underlying generation mechanism. The analysed events show above-average solar wind speeds (except for event 1), negative IMF Bz, and positive By, with a westward convection of FAEs. They are not limited to a certain time sector, with occurrences both between 10:30–11:30 and 21:15–23:15 magnetic local time (MLT). The elevated electron temperature at E-region altitudes and simultaneous increases in ion temperatures at higher altitudes can provide some clues about the origin of FAEs. One possible group of generation mechanisms for the required energy to excite FAEs are Farley–Buneman instabilities, which are streaming instabilities typically occurring at altitudes of 90–120 km . The proposed FAE altitude falls within this region. They become significant when the difference between electron and ion drift speeds exceeds the ion acoustic speed , which is generally the case in geomagnetically disturbed conditions, typically also resulting in aurorae. This would explain why FAEs are observed alongside aurorae. Particularly at high latitudes, these instabilities can result in significant local electron heating. This is consistent with the low-altitude elevated electron temperatures observed during the FAE events, for which Farley–Buneman instabilities are the most likely explanation.

The observed large ion temperatures in the F region around 190 km height are caused by Joule heating from strong electric fields or ion-neutral friction. The measurements are used to estimate the electric field strength below, assuming that E∥=0, i.e. the magnetic field lines, are equipotentials. We neglect the effect of the slightly different magnetic field strengths between 190 km height and the lower E region, and also any differences in the neutral wind between these altitudes. The ion energy balance, neglecting also thermal energy transfer to/from electrons (whose temperatures are generally not enhanced above the E region, especially preceding the FAE occurrence at 18:23 UTC) is Eq. 4: 1Qin=νinNnNe32kBTn-Ti+12miVi-Vn2 Here, Ti and Tn are the ion and neutral temperatures, and Vi and Vn are the ion and neutral drifts, respectively. mi is the mean ion mass, kB is the Boltzmann constant, νin is the ion-neutral collision frequency, and Nn and Ne are the neutral and electron densities. In the steady-state Qin=0, and for the F region, we insert Vi-Vn=E⟂×B/B2 with E⟂ the electric field in the frame of the neutral gas and B the geomagnetic field. We are only interested in the magnitude of E⟂, which can be estimated as follows: 2E⟂/B=3kBTi-Tn/mi. Filtering out elevated ion temperatures above 1500 K, we use the ESR data to estimate a mean background ion temperature in the quiet state of ∼ 950 K for the altitude range of 150–300 km, which should then approximately correspond to the neutral temperature. For mi, we conservatively use 30 amu, corresponding to a mixture between N2+ and O2+, neglecting a contribution by O+. The motivation is that high Ti and large drift difference Vi-Vn probably enhance the relative molecular concentration compared to model values, as the International Reference Ionosphere (IRI) would give it. Using average elevated Ti≈ 3300 K for the altitude range of 150–300 km from the ESR measurement, the estimated lower limit is E⟂≈ 70 mV m-1. This value is far above the threshold for Farley–Buneman instabilities, which is typically around 30 mV m-1 . If molecular ions were assumed to be dominant, it would only further increase the lower limit. It should be noted that this is an approximation, and the filtering for average values is based on somewhat arbitrary choices, but the derived E⟂ is not all that dependent on the inserted Ti and Tn and would exceed the typical limit for Farley–Buneman instabilities by a significant margin regardless of the exact filtering values. The threshold may already be exceeded in the arc, before 18:23 UTC, but Te was perhaps not high enough to excite optical emissions. showed an example with the ESR, where Te reaches temperatures above 3000 K at 100–109 km, which is enough to produce 630.0 nm optical emissions, according to . An open question is whether these instabilities can produce large enough Te increases to excite all observed FAE emissions. showed that Ti increased already above ∼ 125 km, up to about 4000 km. These temperature enhancements are stronger than those observed at auroral oval latitudes over mainland Norway by . This could be because, at the edge of the auroral oval over Svalbard, E fields may be larger than in the auroral zone, or because the ESR is more sensitive than the EISCAT mainland radar was in 1992. If E fields (and associated Te enhancements) are typically larger at Svalbard, this might perhaps explain why FAEs have not been noticed earlier in the auroral zone or along the Scandinavian mainland. Another possible contributing factor could be that auroral all-sky cameras used for scientific purposes are often more limited in pixel resolution compared to the Sony α7S used in the present study, which could reduce the likelihood of unexpectedly identifying small-scale and short-lived features like FAEs.

Whereas specific characteristics for the individually occurring FAEs are hard to identify, category 2 FAE groups with regular spacing clearly suggest a link to wave activity. We tentatively suggest that waves modulate the electric field strength and correspondingly the intensity of Farley–Buneman-induced plasma turbulence and electron heating near the arcs to produce the observed category 2 FAE groups. As these groups show regular and fairly stable distances between the individual FAEs, some kind of monochromatic wave seems to be responsible. describe the modulation of airglow by gravity waves, which is similar to the modulation of category 2 FAE groups, albeit at larger scales. The short distances between FAEs suggests waves with small wavelengths. The estimated FAE drift speed of ∼ 1 km s-1 is much faster than the average ionospheric convection speed of a few hundred metres per second. If category 2 FAEs are indeed modulated by waves, they could propagate with their phase velocity and thus exceed typical convection speeds. Alternatively, the E-field modulation could originate from the magnetosphere. A candidate mechanism is that the shear between the strong flow in the high E field adjacent to the arc and the slower flow in the arc itself leads to a Kelvin–Helmholz instability, whose phase speed would be between the slow and fast flows (see, e.g., ). For E⟂≈ 70 mV m-1 corresponding to 1400 m s-1, the phase speed of Kelvin–Helmholtz waves would be several hundred metres per second, which is roughly the observed value. It is, however, unclear why the auroral arc shows no signature of the modulation and what determines the wavelength of the quasi-periodic FAEs of ∼ 6 km.

To determine a link between FAEs and other aurora-like features like STEVE or the green streaks, and to further analyse FAE characteristics, more events will need to be studied, ideally from multiple locations and with ionospheric plasma measurements. The limited sample size, not necessarily of FAEs, but rather observation nights and ESR data for the present study, limits the conclusions that can be drawn for the underlying generation mechanism. Until these conditions are determined, FAE occurrences will be seemingly random, further complicating a targeted follow-up study.

The focus of the present study is to characterise a new type of aurora-like phenomenon, which we name fragmented aurora-like emissions (FAEs). In summary, the observed FAEs can be grouped into two categories, namely individually occurring FAEs and groups close to auroral arcs with a wave-like structure. All FAEs show a lack of field-aligned extent and seem to generally occur in the shape of an elongated ellipse. The majority of the observed FAEs have a major axis smaller than 20 km (assuming an altitude of ∼ 110 km), with a mean aspect ratio of ∼ 2. Photometer data show distinctly enhanced intensities at the 557.7 nm emission of atomic oxygen for FAEs passing the FOV but no clear FAE signatures at the 427.8 and 630.0 nm wavelengths, of which the latter is not surprising as it would be heavily collisionally quenched at the proposed altitude. A FAE signature is also clearly visible in the ASK1 673.0 nm emission channel of the first positive band system of N2 but not at the 777.4 nm emission of atomic oxygen measured by ASK3, which together sets a range of states with different energies that are excitable by the generation mechanism. The apparent lack of 427.8 and 777.4 nm emissions indicates an upper energy limit between ∼ 8–11 eV which the generation mechanism can produce. The ESR data suggest that FAEs are associated with significantly elevated electron temperatures around 110–120 km, for which Farley–Buneman instabilities are the only known cause at these low altitudes. Simultaneously, increased ion temperatures are visible at altitudes in the F region, which enables us to estimate the strength of the E field. The derived estimate of E⟂≈ 70 mV m-1 exceeds the typical Farley–Buneman threshold of 30 mV m-1. Category 2 FAE groups show a fairly regular and stable spacing and appear to be modulated by some kind of wave.

Open questions are the exact nature of the generation mechanism, such as whether FAEs of categories 1 and 2 are caused by the same mechanism, if category 2 FAEs are indeed modulated by wave activity, and, if so, by what kind of wave, whether they are exclusively a high-latitude phenomenon, and what threshold values of ionospheric parameters are necessary for FAE occurrences.