D-region impact area of energetic particle precipitation during pulsating aurora

Abstract. Ten radars from the Super Dual Auroral Radar Network (SuperDARN) in Antarctica were used to estimate the spatial area over which energetic electron precipitation (EEP) impacts the D-region ionosphere during pulsating aurora (PsA) events. We use an all-sky camera located at Syowa Station to confirm the presence of optical PsA, and then use the SuperDARN radars to detect HF radio attenuation caused by enhanced ionisation in the D-region ionosphere. The HF radio attenuation was identified visually by examining quick-look plots of the background HF radio noise and backscatter power from each radar. 5 The EEP impact area was determined for 74 PsA events. Approximately one third of these events have an EEP impact area that covers at least 12◦ of magnetic latitude, and three quarters cover at least 4◦ of magnetic latitude. At the equatorward edge of the auroral oval, 44% of events have a magnetic local time extent of at least 7 hours, but this reduces to 17% at the poleward edge. We use these results to estimate the average size of the EEP impact area during PsA, which could be used as a model input for determining the impact of PsA-related EEP on the atmospheric chemistry. 10

local time (MLT) contours in Figure 1 relate to the example event described Section 3. The red circle shows the ASC field of view projected to 100 km height. The radar fields of view are shown in black. For simplicity, we show only the near-range field of view of each radar (180-600 km in range), which is the approximate area where the transmitted radiowaves pass through the 95 D-region ionosphere. Each radar's total field of view extends to over 3500 km in range, creating significant overlap between the fields of view which is useful for measuring the F region ionospheric convection (e.g. Nishitani et al., 2019, Figure 1b).
Each SuperDARN radar consists of a linear array of log-periodic or twin-terminated folded dipole antennas which are phased electronically to produce a beam which can be steered in 16 different azimuthal directions. This beam is narrow in azimuth (3.24 • ) but has a wide vertical extent (∼40 • ). The 16 beams are scanned sequentially every minute, with a ∼3 s integration 100 time for each beam.
SuperDARN radars detect coherent backscatter from electron density structures in the E region and F region ionospheres, and also from the ground following reflection in the ionosphere. The southern hemisphere SuperDARN radars were chosen for this study because they generally detect larger amounts of backscatter on the nightside compared to the northern hemisphere radars. This is probably due to the favorable orientation of the geomagnetic field relative to the radar look directions and geographic 105 latitudes. The near-continuous presence of backscatter on the nightside for most Antarctic radars makes it straightforward to identify periods of reduced backscatter power caused by HF radio attenuation. An example of this procedure is given in Section 3. Note that the three mid-latitude SuperDARN radars located in Australia/New Zealand were not included in this study due to their very large spatial separation from the radars in Antarctica, so they are not shown in Figure 1.
The camera located at Syowa Station is a colour digital camera fitted with a 8mm f/2.8 fisheye lens. The imaging season 110 lasts from April to October, and the camera is programmed to capture images automatically whenever the sun is more than 12 • below the horizon. Exposure times are a few seconds, and the image cadence varies from about 6-30s. Daily keogram (quick-look) plots are constructed by taking a magnetic north-south slice through the all-sky image and then placing these slices on a time axis. Similarly, an ewogram can be constructed using a magnetic east-west slice. PsA can be readily identified in these plots as patchy auroral displays (e.g. Jones et al., 2013;Partamies et al., 2017;Yang et al., 2017)  The first column of Figure 2 shows the background noise measurements from each beam of each radar. The noise data have been binned into 10 min intervals and averaged separately in each bin to remove spikes. For most radars there is some variation in the noise levels between different beams, which may arise from real spatial variations in the atmospheric noise production 130 and ionospheric propagation conditions, or differences in beamforming across the radar field of view. For this work we are not interested in the absolute values of the background noise. Instead we look for sudden changes in the background noise over the entire field of view, which indicate that the HF radio noise has been attenuated in response to enhanced D-region ionisation. For the example event, reduced background noise levels during the PsA event are observed for six out of the eight radars shown, as indicated by the black vertical lines.

135
To justify that the reduced background noise measured at the six radar sites was indeed caused by enhanced D-region ionisation, we also study the backscatter power measured by each radar. These measurements are shown in the right column of the ground. The yellow vertical lines on these plots indicate the time periods for which the background noise was attenuated (determined from the left panel). During these time periods, the backscatter returns from all six radars have reduced power or are completely suppressed. This indicates that the radio waves transmitted by the radar have been attenuated. This combination of reduced background noise and reduced backscatter power is strong evidence that the D-region electron density has been enhanced near the radar site (Bland et al., 2018), and we use these two signatures together in our event selection for this study.

145
For the example shown in Figure 2, there are two radars that did not detect a clear attenuation signature during the PsA event.
These are the McMurdo (MCM) and Dome C East (DCE) radars, located at 80 • and 89 • corrected geomagnetic (CGM) latitude respectively. For these radars there is neither a clear decrease in the background noise level, nor a reduction in the backscatter power. Note that the background noise at MCM does decrease close to the event onset time. This, however, is accompanied by an increase in the echo power, which indicates that the transmitted radiowaves were not strongly attenuated in the ionosphere.

150
Therefore, the reduced noise at MCM is probably related to the normal diurnal variation in the background noise at the radar site, rather than a response to energetic particle precipitation.
Based on the information in Figure 2, we can now make a rough estimate of the spatial coverage of the EEP impact area for this PsA event. This area is shown by the grey shading in Figure 1, which is the region bounded by the six radars that detected the event. This area actually represents the minimum EEP impact area for this event. The poleward edge of the EEP impact pair, extending from ∼22-04MLT at the event onset. Since the event was detected at the KER radar site, we can conclude that the equatorward edge of the EEP impact area extends at least as far as 60 • CGM latitude in the morning sector. No data were available from the SAN or FIR radars for this event, so we cannot determine whether the EEP impact area extends to the locations of these radars.
It is worth commenting briefly on the nature of background radio noise measurements from SuperDARN radars. The background noise consists of (1) natural atmospheric radio noise, (2) man-made noise produced by electrical and electronic equip-165 ment near the radar site, and (3) internal noise in the radar receiver system. The short-and long-term variability in the background noise depends on the relative contribution of these three sources to the total noise level. EEP-related attenuation signatures are easiest to identify when the background radio noise exhibits a smooth diurnal variation in the absence of any D-region enhancement. In Figure 2, we observe this smooth noise variation in the DCE, MCM, ZHO, SPS and SYE data, and the sudden reduction in the background noise at the onset of PsA stands out clearly against the slowly-varying background. This diurnal 170 variation in the noise is controlled by the global atmospheric noise production and ionospheric propagation conditions. In contrast, the noise measurements from the Halley (HAL) radar are highly variable from one 10 min time bin to the next, and also between neighbouring beams, which might arise from either internal receiver noise or man-made noise near the radar site. This variability in the noise measurements at Halley makes the visual identification of HF attenuation signatures more difficult. For this radar, the echo loss from 00:30-02:30 UT is clear (right panel of Figure 2), so we rely more heavily on this parameter to 175 determine whether any HF attenuation has occurred. This type of variability in the background noise measurements for some radars is a key reason for adopting qualitative event selection criteria in this study, as some judgment regarding the usability of the data is required in each case.

Event identification for statistical study
To select PsA events for our statistical study of the EEP impact area, we began with a list of 102 optical PsA events identified 180 visually in keogram plots from the all-sky camera at Syowa Station. This same event list was used by Bland et al. (2019) to determine PsA occurrence rates and durations using the SYE SuperDARN radar paired with the all-sky camera data. The optical PsA events were classified as APA, PPA or PA by visually studying the keogram and ewogram plots, and also the 10 scadence all-sky images for each event. The pulsating auroral structures that we used to distinguish between the different types are described in detail by Grono and Donovan (2018)  that was observed. There were only three events during which PA were observed, and in all cases the PA were preceded by APA so they were classified as APA. Note that our PsA classifications apply only to Syowa Station, as we cannot determine the The event start and end times are marked using black/yellow vertical lines.
PsA type from the radar data. Due to this limitation, this study is designed to detect periods of enhanced D-region ionisation that occur simultaneously with different types of optical PsA observed at Syowa Station.
Quick-look plots similar to Figure 2 were produced for each PsA event in the list. For each radar we determined whether the 195 background noise measurements and the echo power had been attenuated based on a visual inspection of the data. Often there was some uncertainty about whether a given radar had (or had not) detected an EEP signature. To capture this uncertainty, the attenuation for each radar was classified as clearly observed/not observed or probably observed/not observed. If we could not determine the presence or absence of an attenuation signature either way for a given event, that radar was excluded from the analysis. For an event to be included in this study, we required that at least one radar in addition to the Syowa Station

205
-A classification of our uncertainty in whether or not any HF attenuation had occurred at each radar site.
In most cases, the onset times for each radar that detected the event are within 30 minutes of the onset times recorded for the SYE radar. For a few events, the onset times at ZHO, KER or SPS are delayed 1-3 hours after the onset at SYE, but there is still considerable temporal overlap between the two sites. This might indicate a latitudinal expansion of the EEP impact area as the event evolves. Similarly, there are three events where the HAL or KER radars detected attenuation more than one hour 210 before the onset at SYE, which might indicate that the particle precipitation began at lower latitudes and later expanded over Syowa Station.
In this study we have used the all-sky camera at Syowa Station to determine the presence of optical PsA. Since HF attenuation will occur in response to any process that enhances the D-region ionisation, we cannot confirm that the attenuation signatures observed by the radars were caused by PsA specifically. For example, the attenuation observed by the SYE and SYS radars in 215 Figure 2 commences at substorm onset at 00:20 UT, about 20 min before optical PsA are visible in the keogram. This is very typical since PsA are frequently observed in the substorm recovery phase, so the onset times determined from the radar data will often be 10-30 minutes earlier than the onset of optical PsA (Bland et al., 2019). To increase the chances that the attenuation signatures at all radar sites are due to PsA-related EEP, we require that the attenuation observed by any individual radar lasts for at least 1 hr to qualify as a positive event identification at that site. Since substorm expansion phases are relatively short 220 (Juusola et al., 2011;Partamies et al., 2013), this criterion should eliminate brief attenuation enhancements during substorm expansion phases that are not accompanied by PsA in the recovery phase.  with all 74 events clearly present. The SYE and SYS radars detected attenuation during all events for which data were available (74 and 63 events respectively). For the SYE radar, 5 of these events had some evidence of HF attenuation but this signature was less clear compared to the other events. For the KER and HAL radars, most of the positive and negative identifications of HF attenuation were uncertain due to the variability of the background noise and lower backscatter occurrence for these radars, similar to Figure 2. The MCM and DCE radars rarely detect any attenuation during the PsA events. These two radars had good 230 data availability, high echo occurrence, and the background noise exhibits smooth diurnal variations, so we conclude that the EEP impact area rarely extends to these latitudes. The lowest latitude radar, FIR, had very limited data availability, so it is not possible to draw conclusions about the EEP response at that location.
To get a general overview of the spatial coverage of the PsA events, it is helpful to first consider only those events that had data available from most of the radars. Therefore, we now take the subset of events with data available from at least five  . EEP impact area for amorphous pulsating aurora (APA) events with data available from at least five radars. The dark grey shading shows the EEP impact area, and the light grey shading shows the region bounded by the radars that did not detect any attenuation. Red symbols are used to indicate our uncertainty in determining whether any attenuation had occurred.
to 74 • magnetic latitude over a 106 • -wide magnetic longitudinal sector. The DCE and MCM radars are not included in this list because the event detection rate at these sites was close to zero. The FIR radar was also excluded due to the low data availability.
In total, there were 17 APA and 10 PPA events with data available from at least five of the six radars listed above, and the 240 EEP impact area for these events is shown in Figures 4 and 5 respectively. For events where the PsA type changed during the event, we grouped the events according to the PsA type observed at the event onset. The events have been sorted and numbered according to the magnetic local time of the event onset at SYE. The light grey shading shows the region bounded by all radars with available data, and then a portion of this area is shaded dark grey to show the region in which attenuation was detected.
The red symbols are used to indicate our uncertainty in whether attenuation occurred or did not occur at each radar site. Since  the event onset times are slightly different for each radar, we use the dark grey shading to represent the total area over which attenuation was observed during the event. The auroral electrojet (AE) index and the planetary K-index (Kp) values at the event onset time are also shown, and we note that there is no obvious correlation between geomagnetic activity and the size of the EEP impact area.
Two APA events and one PPA event were detected by all radars simultaneously (dark grey shading only sparse, we checked all-sky camera data from Syowa Station to determine the spatial coverage of the optical PsA for these events. For all events detected only by the Syowa Station radars, the latitudinal extent of the optical PsA was narrower than the camera field of view. We show an example of this in Figure 6 for APA event #2, where the optical PsA covers about half of the north-south field of view. We found no examples where the east-west extent of the optical PsA was smaller than the camera field of view.

260
The results presented in Figures 4 and 5 show that the EEP impact area associated with PsA events can vary significantly.
compared to 17% at 75 • latitude (SPS-ZHO pair). We also conclude that the optical PsA event reported by Jones et al. (2013) is probably quite rare. For that event, optical PsA were observed over 10 hours of MLT, but our results indicate that a large majority of PsA events have an impact area covering less than 7 hours of MLT. It is possible, however, that the EEP impact area is smaller than the spatial area over which optical PsA are observed. This could be investigated further using the SuperDARN radars and riometers in North America combined with simultaneous observations from the THEMIS all sky cameras.
In terms of magnetic latitude coverage, Grono and Donovan (2020) report that APA occur in the magnetic latitude range 315 56-75 • , whereas PPA and PA occur over slightly narrower ranges of 57-73 • and 59-74 • respectively. They reported that the APA occurrence probability above about 74 • was 15%, and zero for PPA and PA. For our study, observations from the SPS-SYE and ZHO-SYE radar pairs indicate that the EEP impact area regularly extends as far as ∼75-77 • latitude (35% and 24% respectively). Although Grono and Donovan (2020) reported zero occurrence of optical PPA and PA at these latitudes, there are several PPA events in our dataset for which HF attenuation was detected by the ZHO or SPS radars. In particular, events 320 #1, #7 #8, #9 and #10 in Figure 5, which is half of the events shown in that figure. Kp and AE indices for these PPA events are higher than for the other PPA events, which would play a role in determining the size and location of the EEP impact area.
Differences in the auroral oval location between the northern and southern hemispheres may also contribute to the different latitudinal coverage results in this study compared to Grono and Donovan (2020). We also note that the PsA type classification was determined only at Syowa Station, and it is likely that other PsA types were present elsewhere for many events.

325
Based on the results described above, we can make a rough estimate of the average EEP impact area that could be used as input to an atmospheric model such as WACCM (Marsh et al., 2007;Verronen et al., 2016). As described above, the majority of events cover less than 7 hours of MLT. From the SYE, SYS, KER, SAN and HAL pair combinations shown in Figure 7, we determine that roughly half of the PsA events have an instantaneous MLT coverage of at least 2-3 hours at ∼65 • magnetic latitude. Therefore, the average MLT extent at this latitude is probably around 4-5 hours. The average MLT extent is much 330 narrower at ∼75 • magnetic latitude, with the large majority (65-76%) of events covering less than 2-4 hours simultaneously (SYE-SPS and SYE-ZHO pairs). PsA occurrence rates reach a maximum in the early morning sector, at about 04:00MLT (Jones et al., 2011;Bland et al., 2019;Grono and Donovan, 2020), and this result could be used to centre the EEP impact area in a model PsA forcing. The average latitudinal extent of the EEP impact area is difficult to estimate due to the very sparse latitudinal coverage of the SuperDARN radars, and the latitude dependence of PsA occurrence. We have determined 335 that approximately one third of the events cover 12 • of latitude to the west of Syowa Station (SPS-HAL pair), so the median latitudinal extent would be a few degrees narrower. By combining this result with the latitude-dependent optical PsA occurrence rates (Grono and Donovan, 2020, Fig. 5), one could reasonably assume that the average EEP impact area covers the magnetic latitude range of about 62-70 • .
The four most important radars for estimating the EEP impact area in our study are KER, ZHO, SPS and HAL. These radars 340 provide observations from the equatorward and poleward edges of the auroral oval over a wide longitudinal area. Unfortunately, the HAL radar ceased operations in 2015, so it was not possible to extend the dataset to include more recent events from that location. Although the dataset from the other radars could potentially be expanded to include more events, it may be more fruitful to repeat this analysis using the North American SuperDARN radars, which provide additional coverage from a larger southern hemisphere data were used in this study because they generally detect large amounts of backscatter on the nightside, which makes it straightforward to identify periods of reduced backscatter power. We speculate that the high echo occurrence for the southern hemisphere radars is due to the low operating frequencies used and more favourable geometry with the magnetic field for satisfying the aspect-angle condition for ionospheric scatter detection. Also, the southern hemisphere radars generally operate at just one or two frequencies, which makes it easier to identify attenuation signatures in the background radio noise, 350 which is frequency-dependent. With an improved understanding of the background radio noise measured by SuperDARN radars, it may be possible to reliably identify HF attenuation events using only the background noise parameter. This would improve the suitability of the North American SuperDARN radars to estimating the EEP impact area.
Since our event detection procedure is qualitative, we cannot draw conclusions about the magnitude of the D-region electron density enhancements. This may be possible in the future with improved baselining of the background noise parameter (e.g.

355
Berngardt, 2020), and a better understanding of the sensitivity of SuperDARN radars to PsA-related EEP. For the purposes of atmospheric modelling, however, statistical EEP energy spectra from satellite observations (e.g. Tesema et al., 2020b) would probably provide a more accurate estimate the relative impact of different types of PsA than what is possible with the SuperDARN radars.

360
The spatial extent of the EEP impact area during 74 pulsating aurora events has been estimated using observations of HF radio attenuation from ten SuperDARN radars in Antarctica. We defined the EEP impact area as the horizontal region over which PsA-related EEP has sufficient energy to cause a detectable amount of HF attenuation in the D-region ionosphere. This represents the area over which PsA-related EEP may cause an atmospheric chemical response. For 75% of the events studied, the EEP impact area extended over at least 4 • of magnetic latitude, and 36% of events extended over at least 12 • of magnetic 365 latitude. The MLT extent was found to be larger on average at lower latitudes compared to higher latitudes. Based on our results, and earlier work using optical data, we estimated the average EEP impact area for PsA that could be used as input to an atmospheric model such as WACCM. This average EEP impact area extends from about 62-70 • magnetic latitude and covers about 4-5 hours of MLT. We emphasise that the SuperDARN radars have sparse spatial coverage, so these results provide only a rough estimate of the average EEP impact area.

370
Data availability. The SuperDARN data were obtained from the British Antarctic Survey data mirror (https://www.bas.ac.uk/project/superdarn).