A Study of Automatically Detected Flow Channels in the Polar Cap Ionosphere
K. Herlingshaw, L. J. Baddeley, K. Oksavik, D. A. Lorentzen, E. C. Bland
AA Study of Automatically Detected Flow Channels in thePolar Cap Ionosphere
K. Herlingshaw , L. J. Baddeley , K. Oksavik , D. A. Lorentzen ,and E. C. Bland Department of Arctic Geophysics, University Centre in Svalbard, Longyearbyen, Norway, Birkeland Centre for SpaceScience, University of Bergen, Bergen, Norway
Abstract
This paper presents a new algorithm for detecting high-speed flow channels in the polar cap.The algorithm was applied to Super Dual Auroral Radar Network data, specifically to data from the newLongyearbyen radar. This radar is located at 78.2 ◦ N, 16.0 ◦ E geographical coordinates looking north-east,and is therefore at an ideal location to measure flow channels in the high-latitude polar cap. The algorithmdetected >
500 events over 1 year of observations, and within this paper two case studies are considered inmore detail. A flow channel on “old-open field lines” located on the dawn flank was directly driven underquiet conditions over 13 min. This flow channel contributed to a significant fraction (60%) of the crosspolar cap potential and was located on the edge of a polar cap arc. Another case study follows thedevelopment of a flow channel on newly opened field lines within the cusp. This flow channel is aspontaneously driven event forming under strong solar wind driving and is intermittently excited overthe course of almost an hour. As they provide a high fraction of the cross polar cap potential, thesesmall-scale structures are vital for understanding the transport of magnetic flux over the polar cap.
1. Introduction
For southward interplanetary magnetic field (IMF), the high-latitude plasma convection in the ionospherecan be described on the large scale as twin-cell convection, which flows antisunward across the polar capand returns to the dayside at auroral latitudes. The flow is usually assumed to be constant over thousandsof kilometers with typical speeds of several hundreds of meters per second (MacDougall & Jayachandran,2001). However, on a smaller scale (100–500 km), convection within the polar cap is not a uniform, laminarflow but is instead frequently driven by dynamic mesoscale phenomena. These structured flow enhance-ments occur at many locations and are classified under different names within the literature depending ontheir location, speed, size, and duration. Pinnock et al. (1993) reported a class of longitudinally extendedflow channel events (FCEs) within Southern Hemisphere radar data, where the plasma within the flowchannels traveled with speeds of 2–3 km/s within the dayside cusp. These features were suggested to be anionospheric manifestation of flux transfer events (FTEs), which are reconnection events at the dayside mag-netopause with a mean recurrence rate of 7–8 min (Haerendel et al., 1978; Russell & Elphic, 1978, 1979). Theionospheric response to transient dayside reconnection has been observed and named differently dependingon the context in which it was observed. This can introduce confusion as the observations are linked and insome cases can describe the same phenomenon.One of the other radar signatures of dayside reconnection linked to FTEs and FCEs was detected in SuperDual Auroral Radar Network (SuperDARN) data and named pulsed ionospheric flows (PIFs; Provan et al.,1998, 1999). These periodic bursts of antisunward convection, detected initially by the SuperDARN radarlocated in Finland, have a typical recurrence rate of 7–8 min but this can vary between 5 and 12 min. Neudegget al. (2000) identified a statistical link between FTEs and high speed flows in the polar ionosphere (such asPIFs/FCEs) with over 99% confidence by using Equator-S satellite data and SuperDARN data to observe theeffects of FTEs propagating from the magnetosphere to the ionosphere. McWilliams et al. (2000) found theoccurrence rates and repetition rates of PIFs to be very similar to that of FTEs and poleward moving auroralforms, the optical manifestations of FTEs (Denig et al., 1993; Milan et al., 1999, 2000; Sandholt et al., 1990,1993; Thorolfsson et al., 2000).
RESEARCH ARTICLE
Key Points: • Polar cap flow channels can accountfor a substantial amount (40–60%) ofthe cross polar cap potential• Flow channels can form due todayside reconnection or appear onthe edge of polar cap arcs• Magnetic field lines that opened25 min ago can still cause fast flowchannels deep inside the polar cap
Correspondence to:
K. Herlingshaw,[email protected]
Citation:
Herlingshaw, K., Baddeley, L. J.,Oksavik, K., Lorentzen, D. A., &Bland, E. C. (2019). A study ofautomatically detected flow channelsin the polar cap ionosphere.
Journalof Geophysical Research: SpacePhysics , , https://doi.org/10.1029/2019JA026916Received 3 MAY 2019Accepted 19 OCT 2019Accepted article online 30 OCT 2019©2019. The Authors.This is an open access article under theterms of the Creative CommonsAttribution License, which permitsuse, distribution and reproduction inany medium, provided the originalwork is properly cited. HERLINGSHAW ET AL. – Published online 13 NOV 2019 ournal of Geophysical Research: Space Physics
In addition to poleward moving auroral forms, polar cap arcs (PCAs) are also optical features associated withflow channels. PCAs are caused by precipitating electrons that are often accelerated through a field-alignedpotential drop and occur primarily under the influence of a northward IMF and quiet geomagnetic condi-tions. They are known to be associated with flow shears and therefore flow channels, where the potentialdrop across the flow channel is typically ∼
10 kV and can account for ∼ <
10 min),whereas FC 2 occurs on “old-open field lines,” which underwent reconnection 10–30 min earlier (Andalsviket al., 2011; Sandholt & Farrugia, 2009). PIFs occur on newly opened field lines (corresponding to FC 1)in the cusp, while FC 2 is located immediately poleward of the auroral oval on the dawn and dusk flanks(06–09/15–18 magnetic local time (MLT)). FC 2 are 200–300 km wide channels of enhanced antisunwardflow which typically last for 5–10 min. They are attributed to momentum transfer from the high-latitudeand flank boundary layers on the downstream side of the cusp via the C1-C2 cusp currents. The C1-C2 cuspcurrents (Farrugia et al., 2003; Sandholt et al., 2010) form the system responsible for momentum transferfrom the high-latitude and flank boundary layers on the downstream side of the cusp on old-open fieldlines. These channels can either be “directly driven” by southward turnings in the IMF after an appropriatetime delay, or “spontaneous” events occurring during stable periods of southward IMF where the channelis intermittently excited (Sandholt et al., 2010). An IMF B y -induced asymmetry in the location of FC 2 isnoted by Sandholt and Farrugia (2012), as FC 2 is located mainly on the postnoon/dusk (prenoon/dawn)side of the polar cap for IMF B y < ( >
0) conditions. Andalsvik et al. (2011) expands on this framework bydefining a polar cap flow channel as a latitudinally restricted (a few 100 km) regime of enhanced antisun-ward convection > ×
111 km square, which places a lower limit on the size of channels which can be detected.To give further insight into the structure and spacial/temporal evolution of flow channels, we will analyzeindividual, whole field-of-view (FOV) radar scans from the Longyearbyen SuperDARN radar over a periodof 1 year.The selection of a 0.9 km/s velocity magnitude threshold and the definition of a flow channel is discussedin more details in section 3 of the paper. These, along with additional applied criteria, ensure that only fast,well-defined channels embedded within a region of slower, background convection flow are identified (asshown, e.g., in Figure 1). In this paper, the focus will be on two case studies of flow channels, one on thedawn flank and another within the cusp. The properties, formation, evolution and contribution to the crosspolar cap potential will be examined. A future paper will study the statistics of the detected flow channels.
2. Instrumentation
The Super Dual Auroral Network (SuperDARN) is a global network of over 30 high-frequency coher-ent scatter radars designed primarily for studying F region ionospheric plasma (Chisham et al., 2007;Greenwald et al., 1995). During the common mode of operation, each SuperDARN radar in the networkHERLINGSHAW ET AL. 9431 ournal of Geophysical Research: Space Physics Figure 1.
The field-of-view of the Longyearbyen radar in Magnetic Local Time (MLT)/Magnetic Latitude (MLat)coordinates showing the line-of-sight velocity, where blue and red represent flows toward and away from the radar,respectively. A clear channel of enhanced flow toward the radar is visibly embedded within the slower movingbackground flow. steps through a series of azimuthally consecutive beams, separated by ∼ ◦ increments. Each beam is dividedinto 75 range gates of 45 km resolution. The radars are frequency agile (8–20 MHz) and routinely mea-sure the line-of-sight Doppler velocity, spectral width, and backscattered power from magnetic field-alignedionospheric irregularities. These decameter-scale irregularities drift at the bulk E × B drift velocity in the F region ionosphere.In this study, we specifically used scans from the Longyearbyen SuperDARN (LYR) radar, which is locatedat 78.153 ◦ N, 16.074 ◦ E, 472 m altitude and began operations in October 2016. The data used were recorded incommon mode in 2017 on channel A (9.8–9.9 MHz) at 1 min resolution, which is the time taken to complete ascan of all 16 beams. The LYR radar was selected as the first target of the flow channel detection algorithm asit has an ideal position at a high-latitude with a north-east facing field-of-view (shown in Figure 1), coveringa large area of the polar cap and receiving a large amount of backscatter. The radar can theoretically detectbackscatter up to 3,500 km in range but more regularly records data up to 1,500 km with a latitudinal rangeof approximately 76–82 ◦ (magnetic coordinates).Data from SuperDARN radars can be combined to provide maps of the high-latitude ionospheric convectionusing the “map potential” technique (Ruohoniemi & Baker, 1998), in which the electrostatic potential pat-tern is determined as an expansion in spherical harmonic functions. Line-of-sight velocity measurementsfrom all of the radars are gridded and used to determine the values of the spherical harmonic coefficients,while an IMF-driven model is used to constrain the spherical harmonic fit in areas where data coverage issparse or absent. The model used to generate the convection maps within this study is the TS18 statisticalmodel of ionospheric convection (Thomas & Shepherd, 2018). The electrostatic potential pattern usuallyreaches a maximum near dawn and a minimum near dusk. The cross polar cap potential (CPCP), a proxy forthe strength of the ionospheric convection at a given time, can then be calculated as the difference betweenthe maximum and minimum potential. All magnetic coordinates displayed within this paper are altitudeadjusted corrected geomagnetic coordinates (Baker & Wing, 1989; Shepherd, 2014).This study also uses data from the Defense Meteorological Satellite Program (DMSP). Each DMSP satel-lite has a 101 min polar, Sun-synchronous orbit at an altitude of 840 km. Since 2003, the DMSP satelliteshave housed the Special Sensor Ultraviolet Spectrographic Imager (SSUSI), with a global far ultraviolet (UV)HERLINGSHAW ET AL. 9432 ournal of Geophysical Research: Space Physics Figure 2.
Plots demonstrating the analysis of a simulated flow channel. Panel (a) shows a channel of > imager (165–180 nm; Paxton et al., 1992). Each scan of the oval takes ∼
20 min, and the emissions are pro-duced primarily due to precipitating electrons impacting the upper atmosphere. In this paper, observationsin the Lyman-Birge-Hopfield Long (LBHL) wavelength range (165–180 nm) are presented to provide a globalcontext for each event.Optical images recorded by the Sony a7s All Sky Camera (ASC) at the Kjell Henriksen Observatory(78.148 ◦ N, 16.043 ◦ E, 520 m altitude) were also used in this study to provide a detailed view of the time evo-lution of the auroral emissions in the vicinity of the flow channel. The camera is located 1 km from theLongyearbyen radar and uses a fish-eye lens to capture 180 ◦ color images of the sky with a 4 s exposure time.
3. Flow Channel Detection Algorithm
The aim of the newly developed algorithm was to automatically search through the large SuperDARN dataset and locate fast flowing channels embedded within a slower moving background convection. The algo-rithm was applied on the Longyearbyen radar to all 2017 channel A common mode data. This equated to314 days of 1 min resolution scans and in total ∼ ×
45 km range gate along one beam) within the FOV over three consecutive scans. Only the availabledata in each cell over the three scans were averaged together, so the averages in each cell were calculatedfrom between one and three individual velocity values. From this point onward, we define a “scan” to bea grid of beams and range gates, where each “cell” contains an average velocity at that specific beam andrange gate over any available velocity data within three consecutive minutes. A choice of a 3 min average foreach scan maintains a high enough resolution to detect transient channels associated with reconnection.Figure 2a shows a scan of an idealized flow channel plotted as color contours on a beam-gate grid, wherethe red dotted line represents a slice through the beams at a constant range. This idealized case will be usedto illustrate the steps in the algorithm. If there were no fast flows within the grid, then there could be noflow channel, thus a check for velocity magnitudes above 0.9 km/s was applied to the grid. If the searchyielded nothing, this grid was neglected, the start time was shifted by 1 min and a new scan was generated. Ifthere were fast flows within the grid, it was important to determine if they contain structure or if they wereactually noise and not physical. The cell containing the fast flow was then compared to the eight neighboringHERLINGSHAW ET AL. 9433 ournal of Geophysical Research: Space Physics cells, spanning across the adjacent beams and ranges. If scatter was present in all of the neighbors, every cellcontained velocities of the same sign, and their average velocity exceeded 0.9 km/s, then a velocity structureis said to exist and this represents the threshold of detection of the algorithm.Another test was required to eliminate instances where fast flows may exist but with a gradual gradientacross the structure. Cases such as these are likely to be an artifact of the look direction of the radar, perhapsviewing constant flow at an angle which appears to reveal fast flows gradually transitioning to slower flows.A quantitative check of the gradient is then required and is demonstrated in Figure 2b, which shows thevelocity profile over a slice across the beams at a constant range. The gradient of the velocity is calculated bytaking the difference in velocity ( v − v ) in a sliding window across the entire slice. The algorithm searchesfor two sharp gradients of opposing signs, which should be present on the edge cells of the channel at thetransition points between the inside of the fast moving channel and the slower background flows. These twolocations are defined as the “edge” of the channel and it is now possible to refer to the inside and outside ofthe channel with respect to these edges. The edge cells must be of the same sign as the inside of the channel,and there must not be any missing data within the channel. The threshold placed on the gradient is 400 m s − cell − , as this eliminated most of the slowly varying gradients. Additionally, the inside of the channel isexamined to ensure that on average the flow magnitude is above 0.9 km/s, which allows for variation butprevents the detection of more complicated channels that are harder to analyze. There should also be novelocity values exceeding 0.9 km/s outside of the channel as this region should be the slower backgroundflow. To ensure that the background flow exists around the channel, the points immediately outside of theedge cells along the slice are examined. If >
80% of these points are present and there are > ◦ in a circle around the identified cell and the structure was accepted as a flowchannel if any one of these slices satisfied the listed criteria of the gradient and background tests.Another feature shown in Figure 2a is the black oval encompassing the flow channel, where the semimajorand semiminor axes (marked with black straight lines) demarcate the principal component axis of all theidentified cells of the algorithm (gray dots). The principal components of the identified cells were calculatedto estimate the orientation of the channel with respect to the beam direction (semimajor axis) and the widthof the channel (semiminor axis). The ellipse that bounds the flow channel contains 96% of all identified cells.While a variety of velocity magnitudes have been observed inside flow channels (e.g., Andalsvik et al., 2011, v > km/s; Nishimura et al., 2014, v = . km/s; Oksavik et al., 2005, v = . km/s) there exists nodefinition as to how to define the edge of a channel. As such, we tested the detection algorithm with threedifferent velocity gradient thresholds (400, 500, and 600 m s − cell − ) in combination with a variety of velocitymagnitude thresholds inside the channel (from 500 up to 1,000 m/s). To ensure we have fast flows we setthe velocity threshold toward the higher end of the range of velocities observed in a flow channel at v > . km/s . Using this as a velocity threshold and given the spatial resolution of the radar data (45 km range gatesalong the beam), it was decided that a velocity gradient threshold of 400 m s − cell − would ensure that weare indeed observing channel structures with distinct edges embedded within a slower background flow.Using this criteria, the flow surrounding the channel would then be a maximum of 45% of the flow insidethe channel and this change would occur over a small spatial scale (1 cell). In addition, an examination ofthe actual velocity inside the identified channels using this criteria was also undertaken, which showed anaverage channel velocity of ∼ ournal of Geophysical Research: Space Physics Figure 3.
Four different examples of scans where flow channels were identified by the algorithm for the dates of (a) 28February 2017 08:25 UT, (b) 26 October 2017 06:11 UT, (c) 15 April 2017 07:26 UT, and (d) 28 October 2017 04:33 UT. channels were detected in the dayside polar cap between 9 and 14 MLT. There are also events present alongthe dawn and dusk edges of the polar cap and a small number of events on the nightside.This paper will focus on the in-depth analysis of two case studies that were identified by the detection algo-rithm. These two case studies are indicated by the red dots within Figure 4: Case 1 on the dawn flank andCase 2 on the dayside. Case 1 was chosen to investigate flow channels occurring deep within the polar cap,as the processes behind these channels have not yet been fully explored. Case 2 was chosen as the majorityof events were detected in the dayside polar cap, so taking one from this sample allows discussion of thecharacteristics of flow channels on newly opened field lines. This event also persisted for almost an hour,which allows a time series to be examined and for the formation, evolution, and decay of the channel to bestudied. A later paper will explore the statistics of the identified FCs, including event durations, monthlyoccurrence, and the IMF dependencies of FC location.
4. Case Studies ◦ MLAT/7 MLT
The first case study focuses on an example of a flow channel occurring on the dawn flank (79 ◦ MLAT, 7MLT), deep within the polar cap on 2 October 2017 at 01:10 UT. Figure 5 shows solar wind data from theOMNI 1 min resolution data set, which has been time-shifted to the nose of the Earth's bow shock. Overall,the characteristics of the solar wind show average values with no strong solar wind driving ( B z − B y − ∼ , velocity ∼
420 km/s, and pressure ∼ B y alternated between positive and negativevalues. The FC duration, as detected by the algorithm, is indicated by a yellow highlighted section of theHERLINGSHAW ET AL. 9435 ournal of Geophysical Research: Space Physics Figure 4.
The gray dots show the occurrence in Magnetic Local Time(MLT)/Magnetic Latitude (MLat) coordinates of the flow channel center atthe beginning of each event within the study. The two case studies havebeen indicated by red dots. plot from 01:10–01:12 UT. Flow channels occurring on the dawn and duskflanks in the polar cap fall into the FC 2 category and happen on old-openfield lines, which are field lines that have been opened on the dayside10–30 min earlier. In the case of FC 2 on the dawn flank, a positive B y isalso expected (Sandholt & Farrugia, 2012). Searching in the solar winddata for potential FC triggers that match these specifications yields theblue highlighted section between 00:45 and 00:51 UT in Figure 5. Duringthis interval, B z takes a small magnitude southward turning to a maxi-mum of ∼ − B y switches fromnegative to positive, dominating B z with a value of + B y and B z and also their signs are consistent with the Sandholtand Farrugia (2009) framework for FC 2 on the dawn flank on old-openmagnetic field lines.Figure 6a is a plot of data from the SSUSI onboard the F17 DMSP satellite,showing emissions in the LBHL wavelength. On the dawn flank, the UVemissions show a thick band at 70–80 ◦ MLAT associated with the auroraloval and also a thin branch further poleward ( ∼ ◦ MLAT) that is alignedeast-west. This feature is consistent with a PCA, specifically a bendingarc. Bending arcs form under B y -dominated conditions and in most stud-ied cases B z is close to zero (Carter et al., 2015; Kullen et al., 2015). Theymove primarily antisunward, in contrast to other PCAs that move dawn-ward or duskward. The arc may have been imaged on its antisunwardjourney across the polar cap and was observed in the dawn sector due tothe positive B y component. Figure 6b shows a SuperDARN LYR fan plotfor 1:12–1:15 UT, where the FC is clearly visible as an enhanced regionof antisunward flow. DMSP traversed from east to west and passed 70 MLAT on the dawnside at 01:19 UT,which means DMSP SSUSI recorded the region where the SuperDARN FOV shows the flow channel atapproximately 01:12 UT, the same time as the displayed 3 min average. The FC lies in between the auroraloval and the bending arc and the flow on each side of the FC seems to slow or reverse at some ranges. Thevelocity shear associated with the arc is located at the poleward side of the channel. At the equatorwardside, the radar begins to measure the flow reversal region associated with sunward return flow on closedfield lines. The map potential plot of Figure 6c, shows a dominant dusk cell due to the positive B y and theclear signature of the flow channel in the close ranges. Figure 6c also supports the conclusion that the FClocation is antisunward of the convection throat and a comparison with Figure 6a reinforces that the auro-ral oval emissions are in the same location as the reversal of the convection on the equatorward edge of theFC. Unfortunately, DMSP F17 did not pass directly over the flow channel, so cross-track ion drift velocitymeasurements are unavailable for the interval.To investigate the temporal evolution of the flow channel and its associated optical signatures, Figure 7 (left)shows all sky camera data from Longyearbyen plotted beside SuperDARN LYR fan plots over the interval ofinterest (Figure 7, right). On each ASC image, north, west, south, and east are at the top, right, bottom, and Figure 5.
OMNI solar wind data time lagged to the bow shock for Case 1 on 2 October 2017 01:10 UT. The magneticfield data are in Geocentric Solar Magnetospheric (GSM) coordinates. The area highlighted in blue shows the potentialsolar wind trigger of the flow channel, while the yellow highlighted area marks the flow channel duration.
HERLINGSHAW ET AL. 9436 ournal of Geophysical Research: Space Physics
Figure 6.
Plots for Case 1 on 2 October 2017 showing (a) the DMSP F17 pass recorded by the SSUSI instrument in theLBHL wavelength, crossing left to right over the period 1:09–1:19 UT, (b) the SuperDARN line-of-sight velocity scanfrom 1:12–1:15 UT, and (c) the SuperDARN convection map from 1:12–1:14 UT.
HERLINGSHAW ET AL. 9437 ournal of Geophysical Research: Space Physics
Figure 7. (Left column) The all sky camera images during the formation and decay of the flow channel with thecardinal points and look direction of the SuperDARN Longyearbyen radar. (Right column) The SuperDARNLongyearbyen fan plots at the same times as the all sky images, where the green oval shows an approximatefield-of-view of the all sky camera at an assumed altitude of 125 km for the 557.7 nm emission.
HERLINGSHAW ET AL. 9438 ournal of Geophysical Research: Space Physics
Figure 8.
Schematic of the topology of the magnetic field, current systems,and ionospheric features for Case 1 on 2 October 2017. R1 and R2 are theRegions 1 and 2 Birkeland currents, C1 and C2 are the cusp currentsystems, and red arrows marked J p indicated ionospheric Pedersencurrents. The flow channel velocity is depicted by a blue-filled circle, andthe associated electric field by a black arrow. The polar cap arc and auroraloval are shown in green highlighted areas. left, respectively. In the first image at 01:10 UT, a stable, weak arc stretchesfrom north-east to south-west, which has remained in place for roughly30 min, first appearing at 00:40 UT. A second thin, faint arc formed at 1:10UT and is visible close to this stable arc, aligned in the same direction butmore directly over zenith. The arcs abruptly brighten and merge at 01:13UT, aligning north-east to south-west. The intensification of the arcs isshort-lived and begins to fade at 01:17 UT, becoming less structured andfading almost completely by 01:20 UT.The green oval in each panel of Figure 7 (right) shows the approximatefield-of-view of the all sky camera at an assumed altitude of 125 km forthe 557.7 nm emission. Figure 7 (right) shows that the FC is clearly visi-ble at 1:10 UT. This was when the detection algorithm first identified thechannel. However, the three-scan average with the greatest number ofdetected cells, and therefore the peak intensity of the flow channel evo-lution, was at 01:11 UT. There are insufficient background flows for FCdetection with the algorithm by the 01:13 UT scan, although the highflows associated with the FC center still persist. A manual check of the3 min average scans reveals that the flow channel has a duration of 13min. The FC first begins to form at 1:07 UT, peaks at approximately 1:12UT, and then decays in speed and size until there is very little evidence ofthe channel by 01:20 UT. The evolution of the FC and arc are similar induration and intensity, which could suggest that they are a coupled iono-spheric response to a system driver. The FC center is (on average overthe event) located at 79 ◦ MLAT, ∼ ◦ MLAT/10.5 MLT
The majority of the detected FCs occur on the dayside between 9 and 14 MLT (see Figure 4). Therefore, inorder to examine a case representative of this sample, a period of multiple cusp FCs located on average at10.5 MLT (between 05 and 06 UT) is analyzed in the following case study. Figure 9 shows the IMF magneticfield, density, velocity, and pressure from the OMNI data between 4:00 and 6:30 UT on 7 November 2017.The algorithm detected FCs in six intervals indicated by the yellow highlighted sections on all four panels inFigure 9. Figures 9b and 9d show that the solar wind density and pressure were very high during the entireinterval, although Figure 9c shows a slower than average solar wind velocity. Figure 9a shows a discontinuityHERLINGSHAW ET AL. 9439 ournal of Geophysical Research: Space Physics
Figure 9.
OMNI solar wind data time lagged to the bow shock for Case 2 on 7 November 2017 showing (a) IMFstrength where the blue and red lines are the B y and B z components, respectively (b) density (c) velocity and (d)pressure. Yellow highlighted areas show the duration of each flow channel and the solid black lines in panel (a) marktimes of interest that are elaborated on in the text. in the solar wind magnetic field at 4:10 UT (first black vertical line), where there is a 20 nT drop in B z to − B y to − By B y has gradually increasedand stabilized at +
10 nT. Although this period of − B z is favorable for dayside reconnection, it is probablynot steady as B z is fluctuating. At 5 UT (second black line), B z turns and remains constantly southward, atwhich point FCs begin to be detected by the algorithm. Over the next hour, there is strong solar wind drivingwith a constant + B y of 10 nT and a steady − B z between − − B z turns positive at approximately 6:15 UT.Figure 10a shows a SuperDARN convection map where the FC was very fast and wide. The dusk cell domi-nates over the dawn cell, which is consistent with a positive B y . Two regions of enhanced flows can be seenin this plot, a clear FC in the close ranges of the radar (77 ◦ MLAT, 10.5 MLT) and another area of > ◦ MLAT, 15 MLT). Figure 10d shows selected fan plots over the course of the event, whichwill be used in conjunction with the solar wind data to discuss the case study. Visual inspection of the fanHERLINGSHAW ET AL. 9440 ournal of Geophysical Research: Space Physics
Figure 10.
Panels for Case 2 on 7 November 2017 showing (a) the SuperDARN convection map with the LYR fanoverlaid, the DMSP SSUSI data in the LBHL wavelength for (b) 5:03 UT and (c) 6:41 UT, and (d) fan plots ofline-of-sight velocity over the duration of the event from the LYR radar. plots in the lead up to the first successful FC detection reveals that the FC begins forming in the near rangegates at 04:20 UT (not shown), a few minutes after the southward turning of B z at ∼ ournal of Geophysical Research: Space Physics Figure 11.
A time series of Case 2 on 7 November 2017. Panels (a) and (b) show range-time graphs for velocity andspectral width, respectively, for beam 13, in the center of the flow channel. (c) The average velocity of the flow channel(red points) with an error bar of 1 standard deviation. (d) The width of the flow channel, and (e) the flow channelpotential (red points) and total cross polar cap potential (blue points).
HERLINGSHAW ET AL. 9442 ournal of Geophysical Research: Space Physics
The location of the FC within the cusp region suggests that we are observing FC 1 on newly opened fieldlines. The flow channel can be seen to be intermittently excited by the − B z , so therefore falls under the“spontaneously driven” category. B z remains negative and drives reconnection until 6:15 UT. So rather thanthe FC ceasing, the disappearance of the FC from the SuperDARN LYR FOV could be due to the reconnec-tion site moving equatorward due to the enhanced dayside reconnection. This inference is supported by theexpansion of the auroral oval observed by DMSP on two separate passes as seen in Figure 10. The pass at5:03 UT (Figure 10b) shows that at the beginning of the interval, the dayside oval sits at ∼ ◦ MLAT, whileat the next pass at 06:41 UT (Figure 10c) the oval has visibly expanded equatorward to approximately 75 ◦ MLAT, out of the range of the SuperDARN LYR radar FOV.Figure 11 shows a time series over the course of the event, beginning 1.5 hr before the algorithm detectedthe FC and ending 45 min after the last detection. Panels (a) and (b) show range-time plots for beam 13,color coded by velocity and spectral width, respectively. Beam 13 was at the center of the FC, so it is idealfor studying its formation and evolution. The vertical black lines mark the first and last detection times ofthe algorithm during this interval. Figure 11a shows that the algorithm detected the majority of the eventeffectively as the highest velocity flows begin and end close to the black vertical lines. Figure 11b shows ahigh spectral width of up to 400 m/s inside the channel, suggesting turbulent flows. There is an apparentpulsing present in both the velocity and spectral width measurements between 5:00 and 5:35 UT, wherethe channel periodically moves from 800–1,200 km slant range. Figure 11c shows an average velocity (redpoints) at times where the algorithm detected a FC, with an error bar of 1 standard deviation to indicate thespread of velocities inside the channel. Values lower than the velocity threshold are present in the velocityspread. This is because the velocity threshold is applied to the average velocity around a cell and between theedges of the channel. This allows for significant variation in the FC at times, but the average velocity alwaysremains above the threshold of 0.9 km/s. Figure 11c shows that the average velocity generally increasesuntil 5:24 UT from values of 900 to 1,300 m/s. After this point there are too few points to determine a trend.Figure 11d shows a large variation in the FC width over the interval from 190–440 km, which classifies theflow channel as a mesoscale feature. Figure 11e shows the FC potential (red points) and the cross polar cappotential (blue points). The FC potential is 30% of the total cross polar cap potential on average and 46% atthe peak of the event (5:23 UT). The values in Figures 11c–11e are derived from the algorithm outputs (apartfrom the CPCP, which is a SuperDARN data product) and the grayed-out regions on these panels indicateperiods when no FCs were detected by the algorithm.
5. Discussion
The newly developed algorithm described in this paper identifies structured fast flows (over 0.9 km/s) withinthe polar cap. These flows are embedded within a slower background convection flow, with a velocity gra-dient of at least 400 m s − cell − on either side of the flow channel. Our algorithm is the first automatedmethod for detecting FCs in SuperDARN data over all MLTs within the polar cap. The algorithm identified546 events in the Longyearbyen radar data during 2017. The stringent criteria applied with the algorithmensures that only well-defined, fast flow channels were identified, which are moving in the same directionas the background convection. The algorithm was most effective in detecting the FCs at their peak, at thetime of highest flow velocity due to the high-velocity selection criteria. Both short and long duration eventswere identified at a range of MLTs. Detected FCs should be located in the polar cap in a statistical sense, asthe most poleward section of the open-closed field line boundary (dayside cusp region) is statistically locatedat 75 ◦ latitude (Yeoman et al., 2002) and the SuperDARN LYR radar measures a further poleward latitudinalrange of approximately 76–82 ◦ .Two case studies were chosen for further investigation: Case 1 on the dawn flank and Case 2 in the cuspregion. The Case Study 1 event was chosen for further discussion as relatively little is known about flowchannels occurring deep within the polar cap on field lines which have been opened 10–30 min previously.Case Study 2 was chosen as the vast majority of the detected events within the Longyearbyen radar FOVoccurred on the dayside. The event was interesting as it allowed discussion of a dayside event (which makeup the majority of our sample) but was also chosen as it was roughly an hour in duration. This allows ameaningful time series to be examined and for the formation, evolution, and decay of the channel to be dis-cussed. In both case studies, DMSP observations place the auroral oval equatorward of the FC observations,confirming that the FCs occur within the polar cap.HERLINGSHAW ET AL. 9443 ournal of Geophysical Research: Space Physics Case Study 1 on 2 October 2017 shows a short-lived flow channel, lasting 13 min by eye and detected by thealgorithm for 2 min, occurring on the dawn flank. The average velocity of the flow channel was 985 m/s withan average width of 418 km. This yields an electric field value and potential drop across the FC of 49 mV/mand 21 kV, respectively.Case Study 2 on 7 November 2017 is an example of a longer-lived flow channel in the cusp, lasting forapproximately 1 hr. The FC had an average speed of 1.1 km/s, an electric field value of 55 mV/m, and anaverage width of 307 km. The spectral width measurements inside the channel show high values of 400 m/s,suggesting turbulent flows. The average potential across the channel is 17 kV with a peak of 25 kV.
At their peak values, the flow channels accounted for 60% and 46% of the CPCP in Cases 1 and 2, respectively.These are higher percentages than previously observed for FC 2, for example, Andalsvik et al. (2011) (35%)and Sandholt et al. (2010) (25%). Our algorithm gives a more accurate estimate as both previously publishedvalues were observed using DMSP passes, which only give measurements over the satellite trajectory at agiven instance. Our algorithm evaluates the FC in two dimensions (range and beam), sampling the radarFOV at a 1 min resolution. This allows continuous observation of the channel for as long as it remains withinthe radar FOV and sufficient backscatter is present. It is therefore possible to observe the FC over time andobtain average values of the potential. Also, this study does not limit the data to extreme IMF conditions,such as the interplanetary coronal mass ejections typically used by Andalsvik et al. (2011) and Sandholt et al.(2010). Case 1 shows that FCs are occurring for more average values of IMF and an unremarkable, smallmagnitude magnetic field can still generate FCs which account for 60% of the CPCP. FCs have high velocitiesand potentials but are small in geographic area, and will not reproduce as well as large-scale features inconvection map contours. This is due to filtering by the finite spherical harmonic expansion and due to theinfluence from the map potential model. Identifying FCs in the data from the individual radars is thereforeessential to detecting the smaller-scale FCs, and the LYR radar is in an optimal position for the detection ofpolar cap FCs. As polar cap FCs can account for such a significant fraction of the CPCP, it is important tohave radar coverage in the polar cap in order to obtain realistic values of the CPCP. Without the polar capradars, the CPCP would likely be severely underestimated.In both cases, IMF B y is the dominant IMF component. Under these conditions, a magnetic tension withinthe dawn-dusk direction is applied to the newly opened magnetic field lines. The entire convection pat-tern reconfigures on a scale of minutes with a dominant dusk/dawn cell for positive/negative IMF B y in theNorthern Hemisphere (Grocott & Milan, 2014). Reconnection with a B y component then introduces asym-metric loading of magnetic flux into the magnetospheric lobes. As FC comprise a large fraction of the totalCPCP, they are efficient at transmitting this asymmetric loading into the ionospheric convection pattern.This asymmetry can then be reduced when tail reconnection occurs, for example, during substorms, andthe magnetospheric lobes are asymmetrically unloaded (Ohma et al., 2018; Reistad et al., 2018). The algorithm works well at picking up the peak of the FC but does not detect the formation and decay whenthe velocities are below the detection threshold. The case studies were manually inspected after detectionto give insight into the formation and decay processes of the FC. The FC in Case 1 is linked to a PCA,which is likely a bending arc due to the preceding solar wind conditions. The optical emissions potentiallyassociated with the arc and FC are observed in the KHO ASC, aligned in the direction of the SD FOV. Dueto a lack of overlapping fields of view between the instruments used in this case study and the differentobservation parameters (wavelengths and scale sizes), it is not possible to conclusively link the bending arcto the auroral features seen in the ASC. However, the features could be an ionospheric response causedby the same magnetospheric driver due to their similar orientation, duration, and modulation of intensity.These observations support previous work which find FCs occurring on the edges of PCAs (Gabrielse et al.,2018; Zou et al., 2015b).Case 2 shows a FC intermittently excited around the cusp region, strongly driven by a dense, high-pressureIMF, high magnitudes of + B y and a sustained − B y . The − B z persists for over an hour and the FCs are onnewly opened field lines, which makes Case 2 a spontaneously driven FC 1 in the Sandholt framework(Sandholt et al., 2010). There are signs of the structure of the FC forming approximately half an hour beforethe algorithm detects the FC. The velocity however shows a rapid onset in the beam directed along the centerHERLINGSHAW ET AL. 9444 ournal of Geophysical Research: Space Physics of the FC, where speeds jump above the 0.9 km/s threshold and the FC emerges rapidly from the slower flow.Figure 11a shows possible pulsing of the velocity (5:00–5:35 UT) as the channel appears to move slightlybetween 800 and 1,200 km in slant range. This is expected from dayside reconnection phenomena, such asPIFs (Provan et al., 1998), but will not be further analyzed as it is outside the scope of this paper. After thisperiod, the FC seems to stabilize in position from 5:35–6:00 UT and remain at 700–1,100 km slant range.Over the lifetime of the FC, there are high spectral widths of 400 m/s, suggesting turbulent flows within thechannel, which could be structuring at smaller scales than one SuperDARN range gate (45 km). The FC in Case 1, residing on the dawn flank, was 418 km in width, lasted for 13 min (detected by thealgorithm at its peak for 2 min) and was detected 25 min after a small deviation from northward to southwardIMF initiated a reconnection burst with a dominant IMF B y positive component. This width, duration, anddelay time indicate a directly driven FC 2 category on old open flux within the Sandholt and Farrugia (2009)framework. Despite small IMF magnitudes, fast flows are driven deep inside the polar cap, accounting for60% of the CPCP. The FC is observed between a thin, poleward band of emissions and equatorward auroraloval emissions in the DMSP SSUSI data. Due to the B y dominant conditions and as B z is close to zero, theband of emissions most likely falls into the bending arc subclass of PCAs (Carter et al., 2015; Kullen et al.,2015). Bending arcs are located on open field lines (Carter et al., 2015), which further supports the theorythat the flow channel is occurring on old open field lines on the dawn flank, deep within the polar cap. Thiswork builds upon the work of Sandholt and Farrugia (2009), using a similar velocity threshold (0.9 km/s)as compared to the 1 km/s velocity threshold of Andalsvik et al. (2011). For the first time, FC 2 is foundin conjunction with a PCA (specifically, a bending arc) on the dawn flank through combined observationsfrom DMSP, SuperDARN LYR fan plots, and SuperDARN convection maps.
6. Conclusions
A new algorithm was developed to locate flow channels within the polar cap. The algorithm detected 546events over a years' worth of data (2017) from the Longyearbyen SuperDARN radar. Two case studies wereselected for further analysis: Case 1 on the dawn flank and Case 2 in the cusp. The main findings from thesecase studies can be summarized as follows:• The algorithm identified 546 events within the 1 year interval. FCs were observed in the polar cap overmost magnetic local times, but the vast majority were detected in the dayside polar cap.• The FCs comprise high values of the CPCP, peaking at 60% and 46% for Cases 1 and 2, respectively. Thus,polar cap FCs play an important role in flux transfer through the solar wind-magnetosphere-ionospheresystem.• Case 1 confirms that FCs do occur on the edge of PCAs and that fast ionospheric flows can form deepinside the polar cap under small magnitude IMF fields that are B y dominant.• Case 2 shows that fast flows can be driven in the cusp for extended periods with a negative B z com-ponent of the IMF and a high magnitude positive IMF B y . The flow inside these channels is turbulent,exhibiting higher spectral widths for faster flows, which suggests structuring at spatial scales less than oneSuperDARN range gate (45 km).The two case studies provide confidence in the ability of the algorithm to identify FCs in the polar cap. Afuture publication will detail the findings of a statistical study using all 546 identified events. References
Andalsvik, Y. L., Sandholt, P. E., & Farrugia, C. J. (2011). Dayside and nightside contributions to cross-polar cap potential variations: The20 March 2001 ICME case.
Annales Geophysicae , (11), 2189–2201. https://doi.org/10.5194/angeo-29-2189-2011Baker, K. B., & Wing, S. (1989). A new magnetic coordinate system for conjugate studies at high latitudes. Journal of Geophysical Research , (A7), 9139. https://doi.org/10.1029/JA094iA07p09139Birkeland, K. (1908). The Norwegian aurora polaris expedition 1902–1903.Carter, J. A., Milan, S. E., Fear, R. C., Kullen, A., & Hairston, M. R. (2015). Dayside reconnection under interplanetary magnetic field B y -dominated conditions: The formation and movement of bending arcs. Journal of Geophysical Research: Space Physics , , 2967–2978.https://doi.org/10.1002/2014JA020809Chisham, G., Lester, M., Milan, S. E., Freeman, M. P., Bristow, W. A., Grocott, A., et al. (2007). A decade of the Super Dual Auroral RadarNetwork (SuperDARN): Scientific achievements, new techniques and future directions. Surveys in Geophysics , (1), 33–109. https://doi.org/10.1007/s10712-007-9017-8 Acknowledgments
We would like to thank Steve Milan,Jenny Carter, Jade Reidy, and themembers of the Birkeland Centre forSpace Science for useful sciencediscussions. Financial support wasprovided by the Research Council ofNorway under Contract 223252. Solarwind and IMF data are available at theGoddard Space Flight Center SpacePhysics Data Facility (https://cdaweb.sci.gsfc.nasa.gov/index.html/).SuperDARN data were obtainedthrough the SuperDARN website atVirginia Polytechnic Institute andState University (http://vt.superdarn.org/). DMSP SSUSI data are freelyavailable at Johns Hopkins AppliedPhysics Laboratory (https://ssusi.jhuapl.edu/). We thank Larry Paxton(PI of the DMSP SSUSI instrument)and Dag Lorentzen (PI of SuperDARNLongyearbyen) for use of the data.Quick look all sky camera data fromthe Kjell Henriksen Observatory areavailable online (http://kho.unis.no/)and the individual all sky cameraimages used in this paper are availableonline (https://figshare.com/articles/All_Sky_Camera_Images_Svalbard_2_2017/9971246) and were obtained fromthe instrument PI Dag Lorentzen([email protected]).
HERLINGSHAW ET AL. 9445 ournal of Geophysical Research: Space Physics
Denig, W. F., Burke, W. J., Maynard, N. C., Rich, F. J., Jacobsen, B., Sandholt, P. E., et al. (1993). Ionospheric signatures of daysidemagnetopause transients: A case study using satellite and ground measurements.
Journal of Geophysical Research , , 5969–5980. https://doi.org/10.1029/92JA01541Farrugia, C. J., Sandholt, P. E., Maynard, N. C., Torbert, R. B., & Ober, D. M. (2003). Temporal variations in a four-sheet field-alignedcurrent system and associated aurorae as observed during a polar-ground magnetic conjunction in the midmorning sector. Journal ofGeophysical Research , (A6), 1230. https://doi.org/10.1029/2002JA009619Gabrielse, C., Nishimura, Y., Lyons, L., Gallardo-Lacourt, B., Deng, Y., & Donovan, E. (2018). Statistical properties of mesoscale plasmaflows in the nightside high-latitude ionosphere. Journal of Geophysical Research: Space Physics , , 6798–6820. https://doi.org/10.1029/2018JA025440Greenwald, R. A., Baker, K. B., Dudeney, J. R., Pinnock, M., Jones, T. B., Thomas, E. C., et al. (1995). DARN/SuperDARN. Space ScienceReviews , (1-4), 761–796. https://doi.org/10.1007/BF00751350Grocott, A., & Milan, S. E. (2014). The influence of IMF clock angle timescales on the morphology of ionospheric convection. Journal ofGeophysical Research: Space Physics , , 5861–5876. https://doi.org/10.1002/2014JA020136Haerendel, G., Paschmann, G., Sckopke, N., Rosenbauer, H., & Hedgecock, P. C. (1978). The frontside boundary layer of the magnetosphereand the problem of reconnection. Journal of Geophysical Research , , 3195–3216. https://doi.org/10.1029/JA083iA07p03195Iijima, T., & Potemra, T. A. (1976). The amplitude distribution of field-aligned currents at northern high latitudes observed by Triad. Journalof Geophysical Research , , 2165–2174. https://doi.org/10.1029/JA081i013p02165Kullen, A., Fear, R. C., Milan, S. E., Carter, J. A., & Karlsson, T. (2015). The statistical difference between bending arcs and regular polararcs. Journal of Geophysical Research: Space Physics , , 10,443–10,465. https://doi.org/10.1002/2015JA021298MacDougall, J. W., & Jayachandran, P. T. (2001). Polar cap convection relationships with solar wind. Radio Science , , 1869–1880. https://doi.org/10.1029/2001RS001007McWilliams, K. A., Yeoman, T. K., & Provan, G. (2000). A statistical survey of dayside pulsed ionospheric flows as seen by the CUTLASSFinland HF radar. Annales Geophysicae , (4), 445–453. https://doi.org/10.1007/s00585-000-0445-8Milan, S. E., Lester, M., Cowley, S. W. H., & Brittnacher, M. (2000). Convection and auroral response to a southward turning of the IMF:Polar UVI, CUTLASS, and IMAGE signatures of transient magnetic flux transfer at the magnetopause. Journal of Geophysical Research , , 15,741–15,755. https://doi.org/10.1029/2000JA900022Milan, S. E., Lester, M., Cowley, S. W. H., Moen, J., Sandholt, P. E., & Owen, C. J. (1999). Meridian-scanning photometer, coherentHF radar, and magnetometer observations of the cusp: A case study. Annales Geophysicae , (2), 159–172. https://doi.org/10.1007/s00585-999-0159-5Neudegg, D. A., Cowley, S. W., Milan, S. E., Yeoman, T. K., Lester, M., Provan, G., et al. (2000). A survey of magnetopause FTEs andassociated flow bursts in the polar ionosphere. Annales Geophysicae , (4), 416–435. https://doi.org/10.1007/s00585-000-0416-0Nishimura, Y., Lyons, L. R., Zou, Y., Oksavik, K., Moen, J. I., Clausen, L. B., et al. (2014). Day-night coupling by a localized flow channelvisualized by polar cap patch propagation. Geophysical Research Letters , , 3701–3709. https://doi.org/10.1002/2014GL060301Ohma, A., Østgaard, N., Reistad, J. P., Tenfjord, P., Laundal, K. M., Snekvik, K., et al. (2018). Evolution of asymmetrically displacedfootpoints during substorms. Journal of Geophysical Research: Space Physics , , 10,060–10,063. https://doi.org/10.1029/2018JA025869Oksavik, K., Moen, J., Carlson, H. C., Greenwald, R. A., Milan, S. E., Lester, M., et al. (2005). Multi-instrument mapping of the small-scaleflow dynamics related to a cusp auroral transient. Annales Geophysicae , (7), 2657–2670. https://doi.org/10.5194/angeo-23-2657-2005Paxton, L. J., Meng, C.-I., Fountain, G. H., Ogorzalek, B. S., Darlington, E. H., Gary, S. A., et al. (1992). Special sensor ultravioletspectrographic imager: An instrument description. Proc. SPIE 1745, Instrumentation for Planetary and Terrestrial Atmospheric RemoteSensing (Vol. 1745, pp. 2–15). https://doi.org/10.1117/12.60595Pinnock, M., Rodger, A. S., Dudeney, J. R., Baker, K. B., Neweli, P. T., Greenwald, R. A., & Greenspan, M. E. (1993). Observationsof an enhanced convection channel in the cusp ionosphere.
Journal of Geophysical Research , , 3767–3776. https://doi.org/10.1029/92JA01382Provan, G., Yeoman, T. K., & Cowley, S. W. H. (1999). The influence of the IMF B y component on the location of pulsed flows in the daysideionosphere observed by an HF radar. Geophysical Research Letters , , 521–524. https://doi.org/10.1029/1999GL900009Provan, G., Yeoman, T. K., & Milan, S. E. (1998). CUTLASS Finland radar observations of the ionospheric signatures of flux transfer eventsand the resulting plasma flows. Annales Geophysicae , (11), 1411–1422. https://doi.org/10.1007/s00585-998-1411-0Reistad, J. P., Østgaard, N., Laundal, K. M., Ohma, A., Snekvik, K., Tenfjord, P., et al. (2018). Observations of asymmetries in ionosphericreturn flow during different levels of geomagnetic activity. Journal of Geophysical Research: Space Physics , , 4638–4651. https://doi.org/10.1029/2017JA025051Ruohoniemi, J. M., & Baker, K. B. (1998). Large-scale imaging of high-latitude convection with Super Dual Auroral Radar Network HFradar observations. Journal of Geophysical Research , , 20,797–20,811. https://doi.org/10.1029/98JA01288Russell, C., & Elphic, R. (1978). Initial ISEE magnetometer results: Magnetopause observations. Space Science Reviews , (6), 681–715.https://doi.org/10.1007/BF00212619Russell, C. T., & Elphic, R. C. (1979). ISEE observations of flux transfer events at the dayside magnetopause. Geophysical Research Letters , , 33–36. https://doi.org/10.1029/GL006i001p00033Sandholt, P. E., Andalsvik, Y., & Farrugia, C. J. (2010). Polar cap flow channel events: Spontaneous and driven responses. AnnalesGeophysicae , (11), 2015–2025. https://doi.org/10.5194/angeo-28-2015-2010Sandholt, P. E., Dyrland, M., & Farrugia, C. J. (2006). Dayside aurorae and polar arcs under south-east IMF orientation. AnnalesGeophysicae , (12), 3421–3432. https://doi.org/10.5194/angeo-24-3421-2006Sandholt, P. E., & Farrugia, C. J. (2009). Plasma flow channels at the dawn/dusk polar cap boundaries: Momentum transfer on oldopen field lines and the roles of IMF B y and conductivity gradients. Annales Geophysicae , (4), 1527–1554. https://doi.org/10.5194/angeo-27-1527-2009Sandholt, P. E., & Farrugia, C. J. (2012). Plasma flows, Birkeland currents and auroral forms in relation to the Svalgaard-Mansurov effect. Annales Geophysicae , (5), 817–830. https://doi.org/10.5194/angeo-30-817-2012Sandholt, P. E., Lockwood, M., Oguti, T., Cowley, S. W. H., Freeman, K. S. C., Lybekk, B., et al. (1990). Midday auroral breakup events andrelated energy and momentum transfer from the magnetosheath. Journal of Geophysical Research , , 1039. https://doi.org/10.1029/JA095iA02p01039Sandholt, P., Moen, J., Opsvik, D., Denig, W., & Burke, W. (1993). Auroral event sequence at the dayside polar cap boundary:Signature of time-varying solar wind-magnetosphere-ionosphere coupling. Advances in Space Research , (4), 7–15. https://doi.org/10.1016/0273-1177(93)90305-UShepherd, S. G. (2014). Altitude-adjusted corrected geomagnetic coordinates: Definition and functional approximations. Journal ofGeophysical Research: Space Physics , , 7501–7521. https://doi.org/10.1002/2014JA020264 HERLINGSHAW ET AL. 9446 ournal of Geophysical Research: Space Physics
Thomas, E. G., & Shepherd, S. G. (2018). Statistical patterns of ionospheric convection derived from mid-latitude, high-latitude, andpolar SuperDARN HF radar observations.
Journal of Geophysical Research: Space Physics , , 3196–3216. https://doi.org/10.1002/2018JA025280Thorolfsson, A., Cerisier, J.-C., Lockwood, M., Sandholt, P. E., Senior, C., & Lester, M. (2000). Simultaneous optical and radar signaturesof poleward-moving auroral forms. Annales Geophysicae , (9), 1054–1066. https://doi.org/10.1007/s00585-000-1054-2Yeoman, T. K., Hanlon, P. G., & McWilliams, K. A. (2002). Letter to the Editor: A statistical study of the location and motion of the HFradar cusp. Annales Geophysicae , (2), 275–280. https://doi.org/10.5194/angeo-20-275-2002Zou, Y., Nishimura, Y., Lyons, L. R., Donovan, E. F., Shiokawa, K., Ruohoniemi, J. M., et al. (2015a). Polar cap precursor of nightside auroraloval intensifications using polar cap arcs. Journal of Geophysical Research: Space Physics , , 10,698–10,711. https://doi.org/10.1002/2015JA021816Zou, Y., Nishimura, Y., Lyons, L. R., Shiokawa, K., Donovan, E. F., Ruohoniemi, J. M., et al. (2015b). Localized polar cap flow enhancementtracing using airglow patches: Statistical properties, IMF dependence, and contribution to polar cap convection. Journal of GeophysicalResearch: Space Physics , 4064–4078. https://doi.org/10.1002/2014JA020946, 4064–4078. https://doi.org/10.1002/2014JA020946