A detailed study of auroral fragments

A DETAILED STUDY OF AURORAL FRAGMENTS

Master’s thesis for the degree of MASTER OF SCIENCE

SPECIALISATION ASTRONOMY AND SPACE PHYSICS

Uppsala University

The University Centre in Svalbard

Joshua Dreyer

Supervisor: Assoc. Prof. Noora Partamies UNIS Topical reviewer: Dr. Stephan Buchert IRF Uppsala

Examiner: Dr. Andreas Korn Uppsala University

20 June 2019

I

Acknowledgements

First of all, I would like to thank Noora Partamies for the outstanding supervision and constant readiness to answer questions. I cannot imagine a more pleasant supervisor to work with and look forward to working on the publication. A big thanks also to Pål Gunnar Ellingsen for his support over the course of this thesis and Mikko Syrjäsuo for providing some very helpful input.

I most likely would not have written a thesis on this topic without Daniel Whiter, who inspired Noora to come up with the idea in the first place and supplied some very helpful data and insight during the last months. A major thanks for that as well, especially since I really enjoyed working on this topic.

Many thanks also to Stephan Buchert and Andreas Korn for their support from Uppsala, which has always been very uncomplicated and helpful.

Mange tak to Line for pushing me over the last few weeks of the writing process, keeping morale high and checking the final version. Also tusen takk to Henrik for helping with the summary in Norwegian. Finally I would like to thank my family for continuously supporting me over my academic career, I would certainly not have been able to get this far without them.

This thesis is the concluding chapter of my time on Svalbard (for now), a place that has provided me with many of the best experiences of my life. I can say with certainty that the past 1.5 years have been an incredible time. I will miss my Arctic home every day, but as with most things in life, it is a good idea to change things up before it becomes necessary. The last thing I want is to grow tired of Svalbard, and there are always opportunities to return. Thanks to everyone who made the time up there so special, you are way too numerous to list. For now I am looking forward to starting my PhD in Uppsala and will see what the future holds.

Sees snart og takk for turen!

I would like to thank the respective providers and scientific teams of the various instruments below that were used during this thesis for making their data available.

The Kjell Henriksen Observatory (all-sky camera and MSP, UNIS) EISCAT Svalbard Radar

EISCAT is an international association supported by research organisations in China (CRIRP), Finland (SA), Japan (NIPR and ISEE), Norway (NFR), Sweden (VR), and the United Kingdom (UKRI).

ASK instrument

ASK is a collaboration between the University of Southampton and KTH, and is funded by PPARC and the Swedish Research Council.

ACE satellite

I would like to thank the ACE MAG and SWEPAM instrument team and the ACE Science Center for providing the ACE data that was used in this thesis.

DSCOVR satellite

This data was obtained from the NOAA National Centers for Environmental Information website.

II

Abstract

Aurora occurs in various shapes, one of which is the hitherto unreported phenomenon of auroral fragments. For three periods of occurrence of these fragments their properties were studied in detail during this master’s thesis, using mainly ground-based instrumen- tation located near Longyearbyen on Svalbard, Norway. A base dataset was constructed from 103 all-sky camera images, manually marking 305 fragments for further analysis.

This thesis reports and describes the fragment observations during the observed events, including the auroral and geomagnetic context. Fragments generally seem to fall into two categories, the first being singular, apparently randomly distributed fragments, and the second being periodic fragments that occur in groups with a regular spacing close to auroral arcs. A typical fragment has a small horizontal size below 20 km, a short lifetime of less than a minute and shows no field-aligned extent in the emission.

The fragments appear mainly west of zenith (73%) during the three observation nights, whereas their north-south distribution is symmetric around the zenith. Almost all of them exhibit westward drift, the estimated speed for one of the fragments passing the field of view of ASK is ∼ 1 km/s. A spectral signature can be seen in the green auroral wavelength of O at 557.7 nm and red emission line of N

at 673.0 nm, but no emission enhancement was observed in the blue wavelengths. One fragment passing the EISCAT Svalbard radar’s field of view shows a local ion temperature increase in a small altitude range of ∼ 15 km, whereas there is no visible increase in electron density. This could be explained by fragment generation due to locally strong horizontal electric fields. A potential mechanism for this might be electric fields of atmospheric waves superposing with the converging electric fields of auroral arcs created by particle precipitation and the corresponding field-aligned currents. The resulting field would be perpendicular to the magnetic field and the auroral arcs, leading to wave-like density variations of excited plasma close to the arcs. Further study is required to verify this hypothesis and improve the understanding of fragment properties determined from the limited dataset used for this thesis.

KEYWORDS

AURORA FRAGMENTS SVALBARD ALL-SKY CAMERA MSP ASK EISCAT

III

Populærvitenskapelig sammendrag

Nordlysfenomenet, eller aurora, er velkjent på tvers av ulike kulturer og har alltid fascinert og inspirert mennesker. En type enda urapportert nordlys opptrer som små fragmenter med horisontale størrelser hovedsakelig under 20 km. Disse dannes nær større aurorastrukturer som f.eks. buer.

På grunn av mangelen på vitenskapelige publikasjoner om disse nordlysfragmentene, er omfanget av denne oppgaven å gi en definisjon og statistisk studie av deres grunn- leggende egenskaper for videre analyse. Totalt 305 fragmentkandidater er identifisert fra en analyse av 131 himmelfotobilder som ble tatt nær Longyearbyen på Svalbard. Disse kandidatene er deretter statistisk analysert og ytterligere instrumentering brukes til å studere deres egenskaper i detalj.

Fragmentene er funnet å ha korte levetider på generelt mindre enn et minutt, de beveger

seg vestover under de observerte hendelsene og er svært asymmetrisk fordelt vest

for zenith innenfor kameraets synsfelt. De ser heller ikke ut til å vise de feltparallelle

tendensene funnet ved andre nordlysfenomener. Et fragment som passerte synsfeltet til

den kraftige EISCAT Svalbard-radaren, viste ingen signifikante økninger i elektrontetthet,

men heller en økning i ionetemperatur. Dette støtter hypotesen om at fragmentene

genereres av sterke horisontale elektriske felt i stedet for energisk partikkelutfelling, noe

som vil gjøre dem til et unikt nordlysfenomen.

IV Table of Contents

Table of Contents

Acknowledgements I

Abstract II

Populærvitenskapelig sammendrag III

List of Figures VI

List of Tables IX

List of Abbreviations X

1 Introduction 1

2 Theoretical background 2

2.1 The solar wind . . . . 2

2.2 The magnetosphere . . . . 2

2.3 Magnetosphere-ionosphere currents . . . . 3

2.4 The ionosphere . . . . 4

2.5 Aurora . . . . 5

3 Instrumentation 6 3.1 All-sky camera (ASC) . . . . 7

3.2 Meridian-Scanning Photometer (MSP) . . . . 7

3.3 Advanced Composition Explorer (ACE) . . . . 7

3.4 Deep Space Climate Observatory (DSCOVR) . . . . 8

3.5 Auroral Structure and Kinetics (ASK) . . . . 8

3.6 EISCAT Svalbard Radar (ESR) . . . . 9

4 Results 10 4.1 Fragment definition . . . . 10

4.2 Database creation . . . . 11

4.2.1 Marking fragments with Fiji . . . . 11

4.2.2 Astrometry . . . . 13

4.3 Distribution and structure . . . . 14

4.4 Fragment sizes and shapes . . . . 18

4.5 Meridian-Scanning Photometer (MSP) . . . . 21

4.6 ACE and DSCOVR space weather satellites . . . . 24

Table of Contents V

4.7 Auroral Structure and Kinetics (ASK) . . . . 28

4.7.1 ASK video observations . . . . 29

4.8 EISCAT Svalbard radar (ESR) . . . . 29

5 Discussion 31 5.1 Fragment distribution . . . . 31

5.2 Fragment sizes . . . . 31

5.3 Fragment altitudes . . . . 32

5.4 Fragment movement . . . . 32

5.5 Potential generation mechanisms for fragments . . . . 32

6 Conclusion 34 References 35 A Appendices 38 A.1 ASC images . . . . 38

A.2 EISCAT results . . . . 40

VI List of Figures

List of Figures

Figure 2.1: Illustration of the field-aligned (Birkeland) currents connecting the magnetosphere to the ionosphere and the horizontal Pedersen and Hall currents in the ionosphere. Reprinted from Le et al. (2010) with permission. . . . 4 Figure 3.1: Map of the area southeast of Longyearbyen, with instrumentation on

the Breinosa mountain [toposvalbard.npolar.no]. . . . 6 Figure 3.2: View above Kjell-Henriksen Observatory (KHO) and EISCAT Sval-

bard radar (ESR) towards Longyearbyen from the Breinosa mountain on Svalbard. . . . 9 Figure 4.1: Examples for the four fragment groups marked with numbers on an

ASC image taken at 07:55:27 Coordinated Universal Time (UTC) on 2017-12-18 (cropped to the southern half for visualisation purposes, east is left). . . . 12 Figure 4.2: The central 1200 × 1200 pixel selection of the ASC image taken on

2017-12-18 at 08:26:48 used for precise position determination with the website nova.astrometry.net on the left (4 middle squares), with the extracted result star map on the right. The image center is marked with a white square and Polaris is marked with an arrow. . . . 13 Figure 4.3: All 262 marked fragment candidates for the event of 2017-12-18, over-

laid on the first image of the series taken at 07:36:35 UTC (circularly cropped for visualisation purposes). . . . 15 Figure 4.4: All 305 fragment locations in pixel coordinates with histogram distri-

bution and kernel density estimation (KDE). Fragments are shaded according to confidence groups, with darker shades being fragments of higher quality. . . . 16 Figure 4.5: Movement of a group of periodic fragments northward of the main

auroral arc (northwest of zenith) over four successive images taken on 2017-12-18 around 07:49:30 UTC. The images are cropped to 1000 × 500 pixels to make fragments easier identifiable. . . . 17 Figure 4.6: Scatter plot of fragment maximum intensity (in counts on the vertical

axis) against area (in pixels on the horizontal axis) with Pearson

correlation coefficients. A histogram of the variables is plotted on the

outer axes, together with a KDE. . . . 18

List of Figures VII

Figure 4.7: Horizontal distances between pixel centres (vertical axis) for varying elevation angles (horizontal axis) and lens projections (line styles), calculated for ASC images with a size of 2832 × 2832 pixels. . . . 19 Figure 4.8: Length of major and minor axes (in km) of fitted ellipses for each

fragment, assuming an altitude of 110 km. Fragments are shaded according to confidence groups, with darker shades being fragments of higher quality. A histogram of the variables is plotted on the outer axes, together with a KDE. . . . 20 Figure 4.9: MSP keogram of the 427.8 nm, 557.7 nm and 630.0 nm channel (top to

bottom) for the event on 2017-12-18 around 07:50 UTC. Intensities are in Rayleigh and plotted using a logarithmic colour palette for each channel. . . . 21 Figure 4.10: Comparison between ASC image and meridian-scanning photome-

ter (MSP) line plot for the fragment on 2017-12-18 at 07:49:45 UTC, marked with a red line in the line plot. The MSP scan line (1° width) is drawn on the ASC image, zenith is marked with a white rectangle. 22 Figure 4.11: Comparison between ASC image and MSP line plot for the fragment

on 2017-12-18 at 07:54:41 UTC, marked with a red line in the line plot.

The MSP scan line (1° width) is drawn on the ASC image, zenith is marked with a white rectangle. . . . 23 Figure 4.12: Comparison between ASC image and MSP line plot for the fragment

on 2017-12-18 at 07:55:53 UTC, marked with a red line in the line plot.

The MSP scan line (1° width) is drawn on the ASC image, zenith is marked with a white rectangle. . . . 23 Figure 4.13: Comparison of ASC images and MSP line scans for the fragment

moving through the MSP scan line on 2017-12-18 around 07:48:30 UTC. The fragment signatures are marked with vertical lines in the green channel (557.7 nm). The MSP scan line (1° width) is drawn on the ASC images in grey, zenith is marked with a white rectangle. . . 24 Figure 4.14: Solar wind data from the ACE satellite for the event on 2015-12-07

between 18:18–18:28 UTC. The data is sampled in 5 minute means. . 26 Figure 4.15: Solar wind data from the DSCOVR satellite for the event on 2017-

12-18 between 07:36–08:26 UTC. The data is sampled in 5 minute means. . . . 27 Figure 4.16: ASK keogram for the event of 2015-12-07 around 18:23:07 UT in the

upper half. ASK1 measuring the 637.0 nm emission of N

(

P) is visible in the lower middle, the right image shows the ASK3 measurement of 637.0 nm emissions of atomic oxygen, and the left image shows ASK1 in the green/blue channel and ASK3 in the red channel. . . . 28 Figure 4.17: Ion temperature and electron density measurements with the ESR for

the event around 18:25 UT on 2015-12-07. . . . 30

VIII List of Figures

Figure 5.1: Hypothesised generation mechanism for auroral fragments due to a superposition of converging auroral electric fields (in green, perpen- dicular to the magnetic field and the arcs) and the electric field of the atmospheric wave field (in red, also perpendicular to arc and B-field).

Fragments are visualised as green boxes with round corners where the auroral converging electric field aligns with the wave field. The magnetic field direction is into the page. . . . 33 Figure A.1: Pearson correlation coefficients for the variables of the base dataset

(all four confidence groups). This matrix was used to easily spot any significant correlations within the main dataset. . . . 38 Figure A.2: Violin plots of fragment pixel positions split over the four confidence

groups. The outline represents a KDE of the positions. . . . 39 Figure A.3: Fragment circularities on the vertical axis plotted against fragment

area in pixels on the horizontal axis. A histogram with KDE is plotted on the outer axes for both variables. . . . 39 Figure A.4: Electron temperature measurements (Kelvin) in bins of 1 minute

length and 5 km height with the ESR for the event around 18:25 UT on 2015-12-07. . . . 40 Figure A.5: Illustration of terms used in thesis, scaled by occurence rate (created

with wordclouds.com). . . . 40

List of Tables IX

List of Tables

Table 2.1: Observed properties of the solar wind near the Earth orbit at 1 AU according to Kivelson and Russell (1995). . . . 2 Table 2.2: Particle species and lifetimes of excited states for the four emission

wavelengths that were analysed during this thesis (Jones 1974). . . . . 5 Table 4.1: Numbers of fragment candidates split into five confidence groups. . . 12 Table 4.2: Results of astrometric calibration of ASC images using nova.astrometry.net. 13 Table 4.3: Spatial distribution of all 305 fragment candidates relative to zenith in

absolute numbers and percentage of all fragments. . . . 14

X List of Abbreviations

List of Abbreviations

ACE Advanced Composition Explorer ASC all-sky camera

ASK Auroral Structure and Kinetics

AZ azimuth

CSV comma-separated values DEC declination

DSCOVR Deep Space Climate Observatory

EISCAT European Incoherent Scatter Scientific Asso- ciation

EPP energetic particle precipitation ESR EISCAT Svalbard radar

FOV field of view

GSM geocentric solar magnetospheric

GUISDAP Grand Unified Incoherent Scatter Design and Analysis Package

HA hour angle

IMF interplanetary magnetic field KDE kernel density estimation KHO Kjell-Henriksen Observatory MLT magnetic local time

MSP meridian-scanning photometer PDS power density spectrum RA right ascension

RGB red-green-blue

UTC Coordinated Universal Time

1

1 Introduction

The high-latitude phenomenon aurora is known under many names, and comes in various shapes. It has fascinated people since it has been first described and still continues to do so. The processes involved in creating the famous polar lights are complex and were not well-understood until the early 20th century, when the Norwegian scientist Kristian Birkeland began to shape the current scientific consensus on the causes for auroral phenomena. Auroral arcs are probably one of the best-known forms of aurora, but there are numerous distinct patterns and underlying physical processes making up a broad spectrum of the phenomenon referred to as aurora. This thesis is focused on one very specific subset of aurora that seems to not have been studied yet, termed fragments.

These small-scale features are characterised by their limited horizontal size of a few kilometres (mostly below 20 km) and offset from neighbouring auroral structures. Based on their appearance on all-sky camera images, the observed lifetimes are typically less than a minute and fragments show seemingly no field-aligned emission structure. The scope of this thesis is to provide a statistical analysis of 305 potential auroral fragments observed during three nights in Longyearbyen on the Arctic archipelago of Svalbard in the winters of 2015 and 2017. As far as possible, a hypothesis on the underlying mechanisms involved in creating this specific type of aurora shall be constructed based on the available data.

Chapter 2 contains a brief introduction to the theoretical background of aurora and the processes involved in creating auroral phenomena. The instrumentation that was used for this thesis is described in Chapter 3. Results of all used data sources can be found in Chapter 4, including a more in-depth definition of fragments, as well as information on their distribution, sizes, spectral characteristics and underlying plasma properties.

The results are discussed and compared to other known auroral features in Chapter 5.

Finally, the entire thesis is summarized in Chapter 6, with an outlook on the possibilities

for further scientific studies. Additional data can be found in the Appendices after the

bibliography at the end of the thesis.

2 2 Theoretical background

2 Theoretical background

To provide a detailed explanation of all the processes involved in the creation of aurora would be difficult in a full-length book, and certainly impossible in a master’s thesis.

The scope of this chapter is to give a concise overview of the basic mechanisms, starting with the generation of the solar wind and its interaction with the Earth’s magnetosphere.

This results in particle precipitation into the ionosphere, where the excitation of different molecules, atoms and ions produces the optical emissions that characterise auroral phenomena.

2.1 The solar wind

The flow of ionised solar plasma that is accelerated out of the outer layer of the Sun is referred to as solar wind. It carries the solar magnetic field with it across the interplane- tary space in a frozen-in state. The atmospheres and space around planets are affected by the energetic particle flux and pressure that is exerted from the solar wind. Variations of the interplanetary magnetic field (IMF), and speed or density of the solar wind will impact the Earth’s magnetosphere-ionosphere system. Fundamentally, changing solar wind parameters will result in different ionospheric conditions, such as varying auroral phenomena. The observed properties of the solar wind near the orbit of the Earth according to Kivelson and Russell (1995) are summarised below. These values can be considered as long-term averages.

proton density 6.6 cm ⁻

flow speed (nearly radial) 450 km/s magnetic field (induction) 7 nT

Higher proton densities, faster flow speeds and enhanced IMF magnitude (especially the southward component) will result in a compressed magnetosphere and higher energy influx into the magnetosphere-ionosphere system.

2.2 The magnetosphere

The region around the Earth in which the geomagnetic field is dominant against the

IMF is called the magnetosphere. It shields the Earth from the solar wind and thus plays

an important role in creating a habitable planet. At distances over a few Earth radii

from surface it can be almost entirely described by a dipole field model with a dipole

2.3 Magnetosphere-ionosphere currents 3

axis that is tilted from Earth’s rotational axis by ∼ 11° (Kivelson and Russell 1995). The geomagnetic field in its current configuration (which changes over long time periods) points from south to north and meets the IMF at the so-called magnetopause, which is the boundary of the geomagnetic field.

A magnetic reconfiguration by merging of oppositely aligned magnetic field lines is termed magnetic reconnection. This is a fundamental process that connects the solar wind to Earth’s magnetosphere, allowing for an influx of solar wind particles. A basic requirement for magnetic reconnection at the subsolar magnetopause is southwards aligned IMF, or negative B

in geocentric solar magnetospheric (GSM) coordinates, so that the IMF alignment is opposing the northward aligned geomagnetic field. In the GSM coordinate system, the x-axis points from the Earth towards the Sun and the x-z plane contains the Earth’s dipole axis (positive north). On the dayside, the reconnection can lead to a direct influx of solar wind particles into the ionosphere within the region of open field lines connected to the IMF, called the cusp. At the high latitudes beneath the cusp, this results in day-time aurora. The more well-known night-time aurora is the result of energetic particle precipitation from the nightside magnetosphere, through additional acceleration processes within the geomagnetic tail.

Fields lines that are reconnecting with the IMF at the subsolar magnetopause will convect across the polar cap towards the nightside due to the pressure of the solar wind. This process is called the Dungey cycle. The field lines will begin to accumulate within the magnetotail’s plasma sheet, where the decreasing distance between the different magnetic field polarity will eventually result in reconnection of oppositely aligned field lines. Energetic particles will be accelerated towards (and away) from the Earth’s ionosphere, precipitating at high latitudes on the nightside. This process results in so-called substorm aurora, a substorm being an elementary energy transfer process during which the magnetotail is charged and the built-up energy is eventually released into the ionosphere and back into the interplanetary space.

2.3 Magnetosphere-ionosphere currents

The magnetosphere is connected to the ionosphere via the field-aligned Birkeland cur- rents. These appear as two sets, namely the region 1 current sheet at high latitudes and the region 2 current sheet towards the equatorward part of the auroral zone (visualised in Figure 2.1). An electric field across the polar cap is created by the up- and downward flowing Birkeland currents. This leads to the ionospheric Pedersen currents, which flow in the direction of the electric field, and the Hall currents, which flow in the direction of − E × B (Kivelson and Russell 1995). Essentially, the Birkeland currents drive the ionospheric convection.

Within the ionosphere, aurora can be accompanied by parallel electric fields that accel-

erate electrons and thus result in aurora-like emissions which do not originate from

particle precipitation.

4 2 Theoretical background

Figure 2.1: Illustration of the field-aligned (Birkeland) currents connecting the magnetosphere to the ionosphere and the horizontal Pedersen and Hall currents in the ionosphere. Reprinted from Le et al. (2010) with permission.

2.4 The ionosphere

The partially ionised layer of the Earth’s atmosphere is called the ionosphere, typically starting at ∼ 60 km and extending up to more than 1000 km. Incoming solar radiation and energetic particle precipitation ionise the atmospheric gases at these high altitudes.

Due to the low densities, the free electrons can exist for significant periods before recombination. Ionisation levels depend on atmospheric density and available ionising radiation, thus resulting in peaks where the combined effect from these two factors maximises.

The ionosphere consists of several regions, namely the D, E and F layers. The D layer is

the lowest in the ionosphere, extending from ∼ 60 km–90 km. Due to the high neutral

densities at these altitudes, most interactions between neutral and ionised species occur

in the D layer. At ∼ 90–120 km altitude, the E layer shows a more pronounced rise in

electron density during the day than the D layer. It consists mostly of O

and NO

ions

that were created from photodissociation by solar ultraviolet radiation. The fragments

that are studied in this thesis are assumed to be in the E region, which includes the

strongest currents in the ionosphere due to electron drift and ion collisions. Usually split

into F1 and F2 layers, the F layer extends from 150 km upwards with O

becoming the

dominating species above 250 km. It is the only region with significant ionisation levels

2.5 Aurora 5

during night. Both D and E layer rely on the solar radiation for ionisation, which results in a sharp decrease of the electron density after sunset (Kivelson and Russell 1995).

2.5 Aurora

Aurora is a result of energetic particle precipitation (EPP), which is a broad term de- scribing high-energy particles entering the upper atmosphere. These energetic particles ionise and excite the local species in the ionosphere, resulting in emissions that are visible as aurora. If the lifetime of an excited state is longer than the mean collision-free time for that species, the excited particle is likely to collide with another particle. This can result in non-radiative de-excitation in a process called collisional quenching. As a consequence, the corresponding emissions will become rarer with decreasing altitude due to the increasing neutral density. One prominent example is the 630.0 nm emission of atomic oxygen, with a lifetime of 110 s for the excited state. This emission is strongly quenched at altitudes below 150 km, which results in the red emission being visible above 150 km.

The green 557.7 nm emission of atomic oxygen is the strongest auroral emission in the visible range, with a higher excitation energy and much shorter lifetime than the red emission. Typical peak heights are ∼ 120 km for the green and ∼ 200–250 km for the red emission. The other two auroral emissions examined in this thesis originate from molecular nitrogen, with even shorter lifetimes than the 557.7 nm emission.

Optical aurora is usually produced by precipitating electrons with energies between 0.1–15 keV (Rees 1969). Properties of the four main emission wavelengths used for the studies in this thesis are summarised below.

Table 2.2: Particle species and lifetimes of excited states for the four emission wavelengths that were analysed during this thesis (Jones 1974).

particle species emission wavelength lifetime of excited state

N

427.8 nm ∼ 70 ns

O 557.7 nm 0.7 s

O 630.0 nm 110 s

N

673.0 nm 6 µs

Aurora is usually measured in Rayleigh (R), which Baker and Romick (1976) define as:

1 [ R ] _{, 10}

 photons s m

column



6 3 Instrumentation

3 Instrumentation

Most of the instrumentation that was used for this thesis is located at the Breinosa moun- tain ( ∼ 78.15° N, 16.04° E) at an altitude of ∼ 520 m near the arctic town of Longyearbyen on Svalbard. All of the optical instrumentation is located in the Kjell-Henriksen Obser- vatory (KHO) building, including different auroral imagers and a meridian-scanning photometer (MSP). KHO was constructed in 2007 as a replacement for the old auroral observatory in Adventdalen, which outgrew the scientific demand and and was dis- turbed by increasing light pollution from nearby Longyearbyen, among other things.

The powerful incoherent scatter EISCAT Svalbard radar (ESR) is located slightly further down the mountain ( ∼ 430 m altitude) and described in section 3.6. The locations of the KHO and ESR can be seen in Figure 3.1.

Figure 3.1: Map of the area southeast of Longyearbyen, with instrumentation on the

Breinosa mountain [toposvalbard.npolar.no].

3.1 ASC 7

3.1 All-sky camera (ASC)

The base dataset for this thesis was created by analysing ASC images. An ASC is a camera with an ultra wide-angle lens that is able to capture the entire sky. In this case, the model was a Sony α7S mirrorless full-frame camera with a Sigma 8 mm, f/3.5 EX DG circular fisheye lens, which provides a full 180° field of view (FOV). The cropped images have a pixel resolution of 2832 × 2832 pixels with a pixel size of 8.4 µm × 8.4 µm. Colour information consists of red-green-blue (RGB) matrices covering the visible spectrum (The Kjell Henriksen Observatory 2019).

The images were taken using an exposure time of 4 s and an ISO of 16000 in intervals 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, with the camera set to take a picture every 10 s.

Modern mirrorless or digital single-lens all-sky cameras provide a relatively cheap and easy solution to capture high spatial resolution images of auroral events and can be operated over extended periods of time, especially if an electronic shutter is available, which is the case with the Sony α7S. The fisheye lens has the disadvantage of distorting the image towards the edges, which could be removed in post-processing and has to be taken into account when calculating spatial scales using pixel distances.

It should be noted that the ASC images have reversed parity, meaning east is left (top is still north).

3.2 Meridian-Scanning Photometer (MSP)

The MSP consists of five channels measuring the main auroral emissions, among them the three wavelengths 427.8 nm, 557.7 nm and 630.0 nm used for analysis of fragments during this thesis.

The instrument is scanning with a 1° FOV from north to south along the local geomag- netic meridian (31° west of geodetic north) using a rotating mirror. Measurements have a time resolution of 8 seconds, consisting of 4 seconds for a full 360° scan plus another 4 seconds for a background scan. Thus, scanning across the sky (180°) takes 2 seconds.

The background measurement is achieved using a tilting interference filter with narrow band-pass ( ∼ 0.5 nm) for each channel. To remove background noise, the background scan is subtracted from the actual scan. The MSP is absolutely calibrated and provides intensity values in Rayleigh (Chen et al. 2015).

3.3 Advanced Composition Explorer (ACE)

NASA’s Advanced Composition Explorer (ACE) satellite is in a Lissajous orbit around

the L1 Lagrangian point, about 1.5 million km from Earth towards the Sun. It has

delivered real-time data on the composition of the solar wind and the embedded inter-

8 3 Instrumentation

planetary magnetic field since 1997, significantly outliving its planned 5 year mission duration (Stone et al. 1998).

The data that was used in this thesis to analyse the space weather conditions preceding the fragment event on 2015-12-07 was taken from the verified (level 2) dataset. It consists of data produced by the ACE magnetic fields experiment (MAG) and the Solar Wind Electron Proton Alpha Monitor (SWEPAM) in 64 s averages (Zwickl et al. 1998). The parameters that were analysed for this thesis are the IMF components and magnitude from MAG, as well as density and speed from SWEPAM.

3.4 Deep Space Climate Observatory (DSCOVR)

Similar to ACE, the Deep Space Climate Observatory (DSCOVR) satellite is in a Lissajous orbit around L1. Originally scheduled for launch as the Triana satellite in 2003, it re- mained in storage until its launch under the DSCOVR name in February 2015. It replaces ACE as the primary solar wind monitor around L1 and provides space weather as well as Earth observation data, capable of sampling the solar wind velocity distributions and IMF changes at unprecedented temporal resolutions of ∼ 50 samples per second (Vech et al. 2016).

For the event on 2017-17-18, the IMF components and magnitude are provided by the DSCOVR magnetometer, whereas solar wind density and speed are measured by a Faraday cup. In this thesis, level 2 (verified) data in one minute averages is used to study the solar wind conditions ahead of the fragment occurrences.

3.5 Auroral Structure and Kinetics (ASK)

The Auroral Structure and Kinetics (ASK) instrument is a high temporal resolution optical auroral imager, located at the ESR. It consists of three channels with individual filters for selected auroral wavelengths and lenses to adjust FOV (Ashrafi 2007). 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 (D. Whiter 2011).

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 (Goodbody 2014). ASK is pointing towards magnetic zenith and shares part of its observation region with the ESR.

The filters used during the periods of the observed events are 673.0 nm (ASK1), 732.0 nm

(ASK2), 777.4 nm (ASK3). Horizontal FOVs for ASK1–3 respectively are 6.2° 3.1° and

6.2° (KTH Stockholm and University of Southampton 2018).

3.6 EISCAT Svalbard Radar (ESR) 9

3.6 EISCAT Svalbard Radar (ESR)

The European Incoherent Scatter radar network consists of three separate incoherent scatter radar systems in northern Scandinavia and the EISCAT Svalbard Radar (ESR), located on the Breinosa mountain near Longyearbyen at 78.15° N, 16.04° E.

The ESR operates at 500 MHz with a peak transmitting power of 1000 kW and a 25 % duty cycle. Pulse lengths are between 1 us–2 ms, frequency and phase can be modulated. A 42 m fixed parabolic antenna is points to the direction of the magnetic field, and a second 32 m parabolic antenna is fully steerable (Unander 2017). ESR can be operated in various experiment modes, using one or switching between both antennas and modifying range and time resolution to obtain optimal results for different observation goals. Pulses can be coded using phase flips, which drastically increases the range resolution compared to long pulse experiments at the cost of time resolution and computing power requirements.

Beam widths are 0.8° for the 32 m dish and 1.6° for the 42 m dish (Goodbody 2014), which corresponds to ∼ 1.5–3 km horizontal FOV at auroral altitudes of 110 km.

Incoherent scatter radars are able to scatter off the ionospheric electrons directly, which follow the ion motion. After processing, the received scatter signal results in the so-called power density spectrum (PDS). Three parameters are directly obtainable from the PDS:

Electron density N

, ion velocity v

and ratio of electron and ion temperature T

/T

. Various other parameters can be calculated with the Grand Unified Incoherent Scatter Design and Analysis Package (GUISDAP) for Matlab using ionospheric modelling.

Figure 3.2: View above KHO and ESR towards Longyearbyen from the Breinosa

mountain on Svalbard.

10 4 Results

4 Results

The base dataset was gathered from 103 ASC images taken at the KHO near Longyear- byen on Svalbard during three nights. The first event (meaning a time period of fragment occurrence) was observed on 2015-11-07 between 20:15:58 Coordinated Universal Time (UTC) and 20:17:27 UTC, consisting of 4 images. The second event was observed on 2015-12-07 between 18:18:14 UTC and 18:27:36 UTC (20 images). By far the longest, the third event was observed on 2017-12-18 and consists of 79 images taken between 07:36:35 UTC and 08:26:48 UTC. Fragment candidates were identified on 74 of the 103 analysed images.

A multitude of instruments located at or near the KHO provides additional data to further analyse the fragments. The ASC data provides the best overview of fragment occurrence, their locations, sizes and brightness. An analysis of the fragments’ spectral properties using MSP data can be found in section 4.5. Auxiliary instrumentation is used to further investigate their characteristics, such as ASK for a more detailed lifetime analysis and the ESR to study their underlying plasma properties.

Additionally, satellites at the L1 Lagrangian point provide space weather data about the characteristics of the solar wind before and during the events. Space weather conditions for the event nights are described in section 4.6. The analysed events were observed early in the morning or early at night, not during typical substorm time periods around midnight magnetic local time (MLT).

4.1 Fragment definition

A typical quasi-stable auroral arc is characterised by its wide extent horizontally across the sky (kilometres to hundreds of kilometres wide in east-west extent) and vertically extending emissions along the field lines, as well as the rather stable nature. Most arcs will appear, twist, move and then slowly fade over a period of minutes (Partamies et al.

2015).

No publications were found on the phenomenon of auroral fragments, so the definitions

and descriptions in the following sections are based on my visual inspection of the base

dataset. Their horizontal extent is on a scale of a few kilometres (mostly smaller than

20 km), and their lifetimes are predominantly limited to less than a minute. They are

further characterised by a distinct offset from the main auroral features, with a more or

less defined outline and intensity enhancement in the visible spectrum. Generally, the

fragments seem to appear close to bigger auroral features, with an offset that is on the

4.2 Database creation 11

same scale as the fragment size. This is especially true for periodic fragments, and less so for sporadic fragments (described in section 4.3). Fragment intensity enhancements seem to not show the field-aligned extent of auroral curtains. Their name emphasises the size discrepancy to the larger auroral structures, as well as their mostly sporadic and random distribution.

4.2 Database creation

As a first step, the ASC jpg images were visually inspected and all fragment candidates manually detected to create a database. This enables a further statistical analysis of positions, sizes and (to a limited extent) lifetimes and intensities.

4.2.1 Marking fragments with Fiji

The fragment candidates were marked and categorised using the freely available image processing software ImageJ (Rueden et al. 2017). In this case, more specifically using the Fiji (Fiji Is Just ImageJ) distribution, which includes additional software libraries and extended analysis capabilities (Schindelin et al. 2012).

This approach proved to be advantageous in several aspects, enabling easy outlining of fragments after visual identification using the freehand selection tool. An automatic selection via threshold values might seem like the better solution at first, but proved to be not feasible due to the random distribution of the fragments, which means some of them occur in areas with high levels of background aurora. The contrast between the fragments and the background for these cases is too low for a threshold selection, but can still be attempted manually.

Using the built-in ROI (Region Of Interest) manager, the selections can be stored and subsequently reproduced for future analysis. It also provides straightforward export of the all selections into a comma-separated values (CSV) file, which can then be used to create a database. The CSV file not only contains the pixel coordinates of the selections, but also their width, height (referring to horizontal and vertical extent on the images, not in the atmosphere) and area, as well as minimum, mean and maximum intensity values. These intensities are based on the integrated intensity of the RGB channels of the image and since these are produced by a camera that has not undergone absolute intensity calibration. They can only be used to compare images of the same camera to each other. This still allows for some simple analysis, for example size to intensity correlation calculations.

The 331 marked fragment candidates were classified into 5 groups for more options

during the statistical analysis. Group 1 is composed of the most well-defined fragments

with clear borders and strong intensity enhancements, whereas fragments in Group 2

are less clearly offset with softer outlines and lower intensity. The fragments in Group 3

are more speculative, for example those close to zenith or overlaying auroral arcs. Group

4 fragments are not clearly identifiable as fragments, with barely visibly offset from the

12 4 Results

Figure 4.1: Examples for the four fragment groups marked with numbers on an ASC image taken at 07:55:27 UTC on 2017-12-18 (cropped to the southern half for visualisation purposes, east is left). Group 1 and 2 fragments are clear with well-defined outlines. Fragments of Group 3 are more speculative and usually less intense, as well as quite hard to outline. Group 4 contains candidates that are highly speculative, which are not clearly offset from other auroral features and usually only appear on one image.

background, weak intensity enhancements and diffuse structure. If a candidate appears in consecutive images, it is usually easier to identify and will therefore be categorised in a higher confidence group. Fragments in front of other auroral features are naturally hard to outline, and will most likely be in Group 3 or 4.

Generally, fragments close to the arcs can be hard to identify due to the background already being quite bright, and those that are offset are mostly rather easy to spot. A special case are fragments close to the zenith, for which the altitudinal distribution of intensity is not visible, these might just be bottom edge enhancements of an auroral curtain. Lastly, marked candidates of Group 4 that were found to be extremely question- able on repeated screenings of the selections were moved to Group 5. These candidates are discarded and excluded from further analysis. Examples for all four categories are shown in Figure 4.1. The multiple screenings of the selection resulted in the following numbers:

Group 1: 21 candidates (6.3%) Group 2: 55 candidates (16.6%) Group 3: 78 candidates (23.6%) Group 4: 151 candidates (45.6%)

Group 5: 26 candidates (7.9%)

This leaves 305 fragment candidates for the further analysis, with 76 of them being of

high quality, a further 78 of medium quality and 151 candidates with questionable value.

4.2 Database creation 13

4.2.2 Astrometry

The fragment coordinates measured with ImageJ are provided in pixel numbers, but the central pixel of the images does not necessarily correspond to exact zenith. Since the further analysis of the fragment data and comparison with auxiliary datasets requires a precise positioning of the fragments, the ASC images need to be calibrated using astrometry. For this purpose, a centred 1200 × 1200 pixel selection of a mostly aurora- free image was uploaded to the website nova.astrometry.net, which maps the background stars and determines the right ascension (RA) and declination (DEC) values of the central pixel. The image used for the calibration was taken on 2017-12-18 at 08:26:48 UTC, which is towards the end of the longest event. Repeating this calibration with other images did not result in significant variations of the results.

Figure 4.2: The central 1200 × 1200 pixel selection of the ASC image taken on 2017-12-18 at 08:26:48 used for precise position determination with the website

nova.astrometry.net on the left (4 middle squares), with the extracted result star map on the right. The image center is marked with a white square and Polaris is marked with an arrow.

The results of the astrometric position determination are summarised in Table 4.2.

Table 4.2: Results of astrometric calibration of ASC images using nova.astrometry.net.

RA, hms: 15

08

41.629

DEC, dms: + 80° 13 ^′ 05.718 ^′′

Size: 72.3 × 72.3 degrees Pixel scale: 16.59 pixel/degree

Orientation: rotated by 30.4 degrees off true N

The pixel scale is especially useful to determine the offset of fragments from zenith,

whereas DEC needs to be converted to elevation (a in degrees), which can be done via

14 4 Results

equation 4.1. This is required to compare the fragment position on the ASC images with the MSP scans and other auxiliary instrumentation.

sin a = sin (DEC) sin φ + cos (DEC) cos φ cos H (4.1) with DEC : declination, φ : geographical latitude

The hour-angle (H) can be obtained from the right ascension (RA) via H = LST − RA, where LST is the local sidereal time (Duffett-Smith 1989, p. 36). This results in a = 87.85°

for the central pixel of the ASC images, meaning the image centre is 2.15° south of zenith.

The slight offset from zenith can be explained due to camera misalignment and imperfect cropping of the image, with a surplus of black pixels outside the circular sky area on the top part of the image square.

4.3 Distribution and structure

Since the observation duration of the 2017-12-18 event is by far the longest, it also unsurprisingly contains the most fragment candidates (262). The analysis of the event on 2015-11-07 resulted in 4 and the 2015-12-07 event in 39 candidates. This amounts to a total of 305 identified fragment candidates. A spatial distribution for all 262 candidates of the 2017-12-18 event can be found in Figure 4.3, overlaid on the first image of the corresponding series (none of the candidates are actually on this image).

Figure 4.4 shows the pixel coordinates of all 305 fragment candidates (excluding Group 5). Whereas the north-south distribution is nearly symmetric with respect to the zenith, a large asymmetry can be seen for the east-west grouping of fragments. The numbers are summarized in Table 4.3.

Table 4.3: Spatial distribution of all 305 fragment candidates relative to zenith in absolute numbers and percentage of all fragments.

location number of fragments percentage of fragments

north of zenith 154 50.5 %

south of zenith 151 49.5 %

east of zenith 82 26.9 %

west of zenith 223 73.1 %

northwest quadrant 134 43.9 %

northeast quadrant 20 6.6 %

southeast quadrant 62 20.3 %

southwest quadrant 89 29.2 %

The X and Y pixel positions of the 305 candidates split by confidence groups are vi-

sualised in Figure A.2. Whereas the X positions (horizontal on the ASC images) for a

4.3 Distribution and structure 15

Figure 4.3: All 262 marked fragment candidates for the event of 2017-12-18, overlaid on the first image of the series taken at 07:36:35 UTC (circularly cropped for visualisation purposes).

majority of fragments are close to the central X coordinate (1416), the Y positions (vertical on the ASC images) seems rather evenly distributed north and south of zenith. There is a distinct lack of high-quality fragments close to the central Y coordinate (1416).

Fragments can be broadly split into two categories, singular ones with often signifi- cant distance from the closest auroral arc and grouped fragments that show wave-like structure and appear in close distance to the auroral arcs. Both groups seem to show a westward drift on average for the observed events.

Singular fragments show stronger intensities and their maximum size is bigger on

average, but their lifetimes seem to be shorter and their distribution is more random

across the field of view. Typical singular fragments can be seen in Figure 4.1 .

16 4 Results

Figure 4.4: All 305 fragment locations in pixel coordinates with histogram distribution and kernel density estimation (KDE). Fragments are shaded according to confidence groups, with darker shades being fragments of higher quality.

The dashed KDE line is only calculated for fragments of Groups 1 and 2.

Grouped fragments often show clear wave-like (spatially regular) distributions in the

positioning of multiple fragments, with longer lifetimes and a more predictable position

close to the arcs. Fragments of this type seem to follow the movement of structures

in the accompanying arc, in this case almost exclusively westward. A typical example

can be seen in Figure 4.5, showing the positions of one group of fragments in 4 suc-

cessive images. In this case, it seems like the strongest intensity enhancement remains

approximately in the same location, while the fragment group moves westwards.

4.3 Distribution and structure 17

07:49:22 UT

07:49:33 UT

07:49:45 UT

07:49:57 UT

Figure 4.5: Movement of a group of periodic fragments northward of the main auroral

arc (northwest of zenith) over four successive images taken on 2017-12-18

around 07:49:30 UTC. The images are cropped to 1000 × 500 pixels to make

fragments easier identifiable.

18 4 Results

4.4 Fragment sizes and shapes

Figure 4.6 shows the correlation between the fragment area and their maximum inten- sities, based on the brightness calculated from the respective unweighted RGB counts using the formula V = ( R + G + B ) /3.

The mean fragment size is 1175 pixels, most fragment sizes are clustered around this value and very few fragments exceed 2500 pixels in size. The largest fragment by far has an area of 19457 pixels, whereas the smallest is only 83 pixels in size. Likewise, the maximum intensity values are heavily clustered around the mean value of 115 counts, with few bright outliers up to a value of 246 counts.

(a) Scatter plot and correlation of all fragment candidates.

(b) Scatter plot and correlation of Group 1 and 2 fragment candidates.

Figure 4.6: Scatter plot of fragment maximum intensity (in counts on the vertical axis) against area (in pixels on the horizontal axis) with Pearson correlation coefficients. The linear regression is calculated using a robust regression estimate, de-weighting outliers. A histogram of the variables is plotted on the outer axes, together with a kernel density estimation (KDE).

The maximum fragment intensity shows a moderate correlation with the fragment area, with a Pearson coefficient of R

= 0.21. This increases to R

= 0.36 when only fragments of Groups 1 and 2 are used for the analysis.

To determine fragment sizes in kilometres, the pixel sizes for varying elevation need

to be determined. This was done using a script and geometric formulas, which as-

sume an auroral height of 110 km and an equisolid lens projection to calculate the

distance between pixel centres. Equisolid projection for an ideal lens is described

by k = dmax/ sin ( 90/2π/180 ) with k being the number of pixels per radian and

dmax = 1416 for pictures with 2832 pixels width (distance between zenith and horizon in

pixels). The results for an assumed fragment altitude of 110 km can be seen in Figure 4.7.

4.4 Fragment sizes and shapes 19

10 20 30 40 50 60 70 80 90

Elevation angle (deg) 0

0.5 1 1.5 2 2.5

Horizontal distance between pixel centres (km)

Auroral altitude h=110 km, cutoff elevation 10 deg

Figure 4.7: Horizontal distances between pixel centres (vertical axis) for varying elevation angles (horizontal axis) and lens projections (line styles), calculated for ASC images with a size of 2832 × 2832 pixels.

With the determined pixel sizes, fragments can now be measured in kilometres. To provide a more accurate determination of the sizes, the outline selections are fitted with an ellipse in ImageJ, which provides the length of major and minor axis in pixels for each fragment. These are then multiplied with the pixel size in kilometres for the respective fragment elevation. The results are visualised in Figure 4.8, with fragments shaded sequentially according to confidence group. The darkest blue represents fragments of highest quality (Group 1), with sequentially lighter colours for Groups 2–4.

The fragment aspect ratio is given by dividing the axes of the fitted ellipses AR =

[ Major axis ] / [ Minor axis ] , with a mean value of AR = 2.04. Most fragments are fairly

circular, with a mean circularity value of c = 0.705 (c = 1 being perfectly circular),

which is determined using the formula c = 4π · [ Area ] / [ Perimeter ]

. The circularity

distribution plotted against fragment area can be seen in Figure A.3.

20 4 Results

Figure 4.8: Length of major and minor axes (in km) of fitted ellipses for each fragment, assuming an altitude of 110 km. Fragments are shaded according to

confidence groups, with darker shades being fragments of higher quality. A histogram of the variables is plotted on the outer axes, together with a KDE.

The dashed KDE line is only calculated for fragments of Groups 1 and 2.

4.5 Meridian-Scanning Photometer (MSP) 21

4.5 Meridian-Scanning Photometer (MSP)

Due to being absolutely calibrated every year, the MSP provides the absolute emis- sion intensities in Rayleigh for the three main auroral lines at 427.8 nm, 557.7 nm and 630.0 nm, further referred to as the blue, green and red channel. These measurements for these three channels are visualised in Figure 4.9

Due to the very limited lifetime of the fragments, the collisionally quenched emissions at 630.0 nm (state lifetime of 110 s) will either not show fragments at all, or only ^smeared over the temporal axis. For this reason, the red channel is unlikely to provide meaningful data for the study of these fragments. Emissions in the blue channel at 427.8 nm and green channel at 557.7 nm on the other hand are almost instantaneous. Based on the visual inspection of the ASC images, fragments are expected to be clearly visible in the green wavelengths, whereas blue emissions are not identifiable on the images. The assumption that the fragments are occurring in a fairly small altitudinal extent around 110 km would explain the dominating emissions in the green, since this is around the peak emission altitude for 557.7 nm emissions from atomic oxygen.

Figure 4.9: MSP keogram of the 427.8 nm, 557.7 nm and 630.0 nm channel (top to bottom) for the event on 2017-12-18 around 07:50 UTC. Intensities are in Rayleigh and plotted using a logarithmic colour palette for each channel.

There are some fragments visible in the green channel of Figure 4.9, indicated by black

arrows. The arc starting around 07:54 UTC is accompanied by fragments southward of

the arc, which are distinctively spatially offset. The blue channel for N

in Figure 4.9

does not seem to show fragment signatures, which will be further analysed in the

individual fragment emission profiles below. Using fragment candidates identified from

the ASC images as an indicator, the precise time for a suitable line plot containing an

emission signature of a fragment can be identified. Four examples of fragment signatures

in the MSP line plot compared to the respective ASC images are shown below. The

22 4 Results

image centres are marked with a white square and the MSP scan line is overlaid with a gap at the fragment location. The MSP scan angle runs from 0° (magnetic north) to 180°

(magnetic south). The emission order for the line plots from top to bottom is blue, green and then red.

(a) ASC image at 07:49:45 UTC.

0 100 200

100 0 200 300 400

0 20 40 60 80 100 120 140 160 180

500 0 1000 1500

MSP scanning angle [degrees]

Emission intensity [Rayleigh] 427.8 nm 557.7 nm 630.0 nm

Emission intensities at 2017-12-18, 07:49:45 UTC

(b) Line plot over the meridian at 07:49:45 UTC.

Figure 4.10: Comparison between ASC image and MSP line plot for the fragment on 2017-12-18 at 07:49:45 UTC, marked with a red line in the line plot. The MSP scan line (1° width) is drawn on the ASC image, zenith is marked with a white rectangle.

The fragment in Figure 4.10a can be seen as part of a group of periodic fragments northward of the auroral arc, around 21° north of zenith. As seen on the MSP line scan, the main auroral arc shows an intensity of 282 R in the 557.7 nm channel, which drops off to 143 R towards the northern edge of the arc and then rises abruptly to 388 R for the fragment signature in a narrow peak (marked with a red line). This peak is only visible in the green channel.

The large fragment around 132° scan angle is visible in Figure 4.11a, accompanied by a strong curtain-like feature northward of the fragment with a maximum intensity of 5434 R. There is a clear gap in intensity visible on the MSP line scan (Figure 4.11b). This supports the hypothesis that the feature is indeed a fragment and not merely the bottom end of an auroral curtain. The strong intensity enhancement for the green emission line is very distinct at 3357 R (red line), whereas there is no enhancement visible in the blue channel.

The spatial offset and intensity enhancement is less pronounced in Figure 4.12, with very low Rayleigh values for the fragment in the green channel. The broad peak from ∼ 110 to 125° scan angle in Figure 4.12b shows the diffuse patch of aurora southward of the main arc close to zenith and northward of the fragment, whereas the fragment itself is visible as a smaller intensity peak at 131° scanning angle with an emission intensity of 520 R (red line). The fragment intensity is ∼ 25% of the maximum emission intensity of 2058 R for the auroral arc close to zenith.

Finally, Figure 4.13 shows the progression of a fragment moving through the MSP scan

4.5 Meridian-Scanning Photometer (MSP) 23

(a) ASC image at 07:54:40 UTC.

0 500 1000

0 2500 5000

0 20 40 60 80 100 120 140 160 180

1000 0 2000 3000

MSP scanning angle [degrees]

Emission intensity [Rayleigh] 427.8 nm 557.7 nm 630.0 nm

Emission intensities at 2017-12-18, 07:54:41 UTC

(b) Line plot over the meridian at 07:54:41 UTC.

Figure 4.11: Comparison between ASC image and MSP line plot for the fragment on 2017-12-18 at 07:54:41 UTC, marked with a red line in the line plot. The MSP scan line (1° width) is drawn on the ASC image, zenith is marked with a white rectangle.

(a) ASC image at 07:55:51 UTC.

100 0 200 300 400

500 0 1000 1500 2000

0 20 40 60 80 100 120 140 160 180

500 0 1000 1500 2000

MSP scanning angle [degrees]

Emission intensity [Rayleigh] 427.8 nm 557.7 nm 630.0 nm

Emission intensities at 2017-12-18, 07:55:53 UTC

(b) Line plot over the meridian at 07:55:53 UTC.

Figure 4.12: Comparison between ASC image and MSP line plot for the fragment on 2017-12-18 at 07:55:53 UTC, marked with a red line in the line plot. The MSP scan line (1° width) is drawn on the ASC image, zenith is marked with a white rectangle.

line and fading away between 07:48:25–07:48:49 UTC. The fragment is clearly visible in

the ASC images, disappearing between the 07:48:35 and 07:48:47 UTC. Relatively strong

intensity enhancements up to 2500 R are visible in the green channel of Figure 4.13d,

marked with vertical lines. At the same time, there are no enhancements visible in the

blue or red channel. It should be noted that this fragment is located close to zenith and

therefore the field-aligned emission structure is not visible on the ASC images. The

fragment might just be the bottom-edge enhancement of a standard auroral curtain in

this case.

24 4 Results

(a) ASC image at 07:48:23 UTC.

(b) ASC image at 07:48:35 UTC.

(c) ASC image at 07:48:47 UTC.

(d) Stacked MSP line scans at 07:48:25 UTC (blue), 07:48:33 UTC (orange), 07:48:41 UTC (green) and

07:48:49 UTC (red).

Figure 4.13: Comparison of ASC images and MSP line scans for the fragment moving through the MSP scan line on 2017-12-18 around 07:48:30 UTC. The fragment signatures are marked with vertical lines in the green channel (557.7 nm). The MSP scan line (1° width) is drawn on the ASC images in grey, zenith is marked with a white rectangle.

All four analysed fragments show intensity enhancements in the green without notice- able increases in the blue or red channel. Due to the collisional quenching, this was expected for the 630.0 nm channel. The fact that there are no enhancements in the blue 427.8 nm channel will be discussed further in chapter 5. All fragment intensity peaks are rather well-defined and clearly offset by an intensity gap from the closest auroral curtain structure.

4.6 ACE and DSCOVR space weather satellites

To analyse the underlying space weather conditions for the fragment events, satellite data from the Sun-Earth L1 Lagrangian point can be used, which can be obtained from the ACE satellite for the event on 2015-12-07 between 18:18 UTC and 18:28 UTC.

Particularly the proton density and velocity, as well as the components of the in-

4.6 ACE and DSCOVR space weather satellites 25

terplanetary magnetic field are of interest for auroral physics. Due to the distance between L1 and the magnetopause ( ∼ 1.5 million km), a travel time has to be esti- mated. A quick calculation using a solar wind proton velocity of 640 km/s results in: ( 1.5 million km ) / ( 640 km/s ) = 2344 s = 39 min. Some additional time needs to be added for the travel time between magnetopause and ionosphere, dependent on the magnetosphere-ionosphere coupling involved for the respective time of day. Since this fragment event was observed during the evening, a significant convection time across the polar cap is required, which is hard to estimate accurately. In total, it is likely that the solar wind data needs to be lagged by 1–2 hours, so the period of interest is approximately 16:20–17:20 UTC.

Figure 4.14a shows an average proton speed of ∼ 640 km/s, with a brief spike up to 680 km/s at 16:55 UTC, before dropping off slightly. These speeds are considered high compared to normal values around 450 km/s.

The observed proton densities are quite low, remaining quite stable around ∼ 2 cm ⁻

. Usually, the components of the interplanetary magnetic field in GSM coordinates as seen in Figure 4.14b are of particular interest for auroral studies due to their significant effect on the reconnection properties at the magnetopause and subsequently the ionospheric distribution of aurora. B

shows a brief rise above 0 nT around 16:45 UTC before dropping to almost − 5 nT around 17:05 UTC. For the same period, B

remains stable between 2.5–5 nT. B

is generally less relevant and is rather stable around − 2.5 nT with a brief spike to 0 nT around 16:55 UTC.

For the event on 2017-12-18 the DSCOVR satellite provides the relevant space weather data, which is visualised in Figure 4.15. The fragment occurrences were observed between 07:36 and 08:26 UTC. Using the same rough estimation as above, the relevant solar wind time period is approximately 05:30–07:30 UTC. Figure 4.15a shows high proton speeds around 620 ± 20 km/s and densities between 7–8 cm ⁻

.

The magnetic field components are plotted in Figure 4.15b. Notably, B

remains negative for the entire period (with one brief spike to 0 nT) and B

is rather stable between 5–7.5 nT. B

is positive for the entire period after 05:40 UTC between 1–9 nT.

Similarities between the two event nights are negative B

for almost the entire period

of interest and a rather stable positive B

. While the proton density is below average

for the 2015 event and above average for the 2017 event (Table 2.1), the proton speed is

quite high for both. This would indicate an efficient transfer of energy from the solar

wind to the magnetosphere-ionosphere system during both events.

26 4 Results

(a) Solar wind proton speed (km/s) and density (cm

), sampled in 5 minute means.

(b) Solar wind magnetic field components (nT) in GSM coordinates, sampled in 5 minute means.

Figure 4.14: Solar wind data from the ACE satellite for the event on 2015-12-07 between

18:18–18:28 UTC. The data is sampled in 5 minute means.

4.6 ACE and DSCOVR space weather satellites 27

(a) Solar wind proton speed (km/s) and density (cm

), sampled in 5 minute means.

(b) Solar wind magnetic field components (nT) in GSM coordinates, sampled in 5 minute means.

Figure 4.15: Solar wind data from the DSCOVR satellite for the event on 2017-12-18

between 07:36–08:26 UTC. The data is sampled in 5 minute means.

28 4 Results

4.7 Auroral Structure and Kinetics (ASK)

Five marked fragments of high to medium quality (Group 1–3) moved through the FOV of ASK around 18:23 UTC on 2015-12-07. One of them is clearly visible in Figure 4.16, slightly offset from the main auroral arc. The upper part of Figure 4.16 is the keogram of the ASK1 and ASK3 channels for the approximately 60 s around 18:23 UTC, whereas the three lower parts show the different ASK channels.

Figure 4.16: ASK keogram for the event of 2015-12-07 around 18:23:07 UT in the upper half. ASK1 measuring the 637.0 nm emission of N

is visible in the lower middle, the right image shows the ASK3 measurement of 637.0 nm emissions of atomic oxygen, and the left image shows ASK1 in the

green/blue channel and ASK3 in the red channel (private communication, Daniel Whiter, 2019).

When viewing successive ASK images, fragments can be seen to form and vanish within

a few seconds or less. ASK provides high frame rate observations of the aurora, which in

turn means it is likely that the majority of the fragments that are visible in the ASK data

were not actually marked for the base dataset, since they appeared and disappeared

within the time span between two consecutive ASC images. The ASK images show

a high number of small fragments in periodic grouping drifting westwards along the

wavering main auroral arc. This is especially visible when observing a continuous video

of the ASK images.