Comparison of the shock arrival times for Earth-directed ICMEs provided by the WSA-Enlil+Cone model and in-situ observations at L1: A Case Study

Examensarbete 15 hp Maj 2016

Comparison of the shock arrival times for Earth-directed ICMEs provided by the WSA-Enlil+Cone model and in-situ observations at L1: A Case Study

Anita Linnéa Elisabeth Werner

Kandidatprogrammet i Fysik

Sammanfattning

Denna fallstudie undersöker överensstämmelsen mellan modell och data för tre interplanetära chock- vågor, som kunde detekteras vid jordens Lagrangepunkt 1, och som orsakade geomagnetiska stor- mar av måttlig till kraftig styrka. Vi använder oss av tidigare genomförda körningar av den sam- mansatta WSA-Enlil+Cone modellen, som avbildar fortplantningen av temporära störningar med ursprung i solens korona, såsom koronamassutkastningar, ut i heliosfären. Modellen gjordes till- gänglig av Community Coordinated Modeling Center (CCMC) och datan inhämtades från OMNI.

Kod skriven i MATLAB nyttjades för att göra en jämförelse mellan modell och faktiska mät- ningar av det interplanetära magnetfältet samt solvindens hastighet, densitet och temperatur. Utöver detta, beräknas också skillnaden mellan förväntad och faktisk ankomsttid av respektive interplan- etär chock, och tidsperioder med en temperatursänkning utöver det normala identifieras.

Vi finner en omfattande avvikelse mellan modell och data, i synnerhet för de fall där på varandra

följande koronamassutkastningar förväntas interagera eller rent av slås ihop samt för uppskattnin-

gen av den omgivande solvindens egenskaper och det interplanetära fältet under pågående geo-

magnetisk störning. Interaktionen mellan koronamassutkastningar och närliggande ko-roterande

interaktionsregioner har ej heller återskapats väl av modellen ifråga. Slutligen ger vi förslag på

möjliga, framtida åtgärder som kan bör tas i åtanke vid konstruerandet av framtida versioner av

nämnda modell, liksom för den allmänna förståelsen för rymdvädrets inverkan på Jorden.

Acknowledgements

I would like to thank Dr. Andris Vaivads, Prof. Hermann Opgenoorth and Dr. Emiliya Yordanova for granting my wish to carry out my final degree project at the Institute of Space Physics at Upp- sala, and more so, enabling me to immerse myself within the research field that I love most: CMEs.

Without their encouragement and relentless support, I would not have achieved half of the under-

standing I hold today of ICME in-situ signatures, space weather modeling and the appropriate data

processing. I am so, so eager to learn more.

Contents

1 Introduction 1

2 Review 2

2.1 CMEs . . . . 2

2.1.1 Definition . . . . 2

2.1.2 Initiation . . . . 2

2.1.3 Kinematic properties . . . . 3

2.1.4 Halo CMEs . . . . 4

2.2 ICMEs . . . . 5

2.2.1 The solar wind . . . . 5

2.2.2 IP shocks . . . . 6

2.2.3 In-situ signatures . . . . 8

2.2.4 Structure: Sheath . . . . 8

2.2.5 Structure: Magnetic cloud . . . . 10

2.2.6 Structure: T

depression . . . . 11

3 Data and Method 11 3.1 The WSA-ENLIL+Cone Model . . . . 13

3.2 Geometric triangulation . . . . 14

4 Results 16 4.1 The 2012 January 18 event . . . . 17

4.2 The 2012 January 19 event . . . . 17

4.3 The 2012 January 22 IP shock . . . . 18

4.4 The 2012 January 23 event . . . . 20

4.5 The 2012 January 24 IP shock . . . . 21

4.6 The 2015 March 15 event . . . . 22

4.7 The 2015 March 17 IP shock . . . . 23

5 Discussion 25 5.1 2012 January 22 . . . . 25

5.2 2012 January 24 . . . . 26

5.3 2015 March 17 . . . . 26

5.4 General analysis . . . . 27

6 Conclusions 29

7 Outlook 29

8 Bibliography

1 Introduction

The first documented attempt at a definition of coronal mass ejections (CMEs) was proposed in the 1930s (Chapman and Ferraro, 1931), then simply described as the ejection of ionized matter from the Sun and, potentially, the long-sought cause of aurora on Earth. The first observational discovery of a CME in modern time can be attributed Hansen with colleagues (Hansen et al., 1971), who in August 1971 reported the findings of a "coronal disturbance", observed through a white-light coronagraph as an intense, short-lived brightening of the Sun’s eastern limb. In 1974, during a time period of 118 days, a staggering amount of 39 instances of transient mass ejections from the solar corona was detected with the use of a white-light coronagraph onboard the space station Skylab (Gosling et al., 1974). This statistical study managed to confirm a great deal of the existing theories regarding the origin of CMEs and its possible connection to other transient solar events, such as solar flares and prominence eruptions.

Since then, the research field concerning the CME phenomenon, including but not limited to its initiation, eruption, propagation, interaction and association with other transient solar phenomenon, contribution to space weather and the modeling of all those effects has received increased attention.

There is more than 18,000 refereed articles on the matter of CMEs (Astrophysics Data System, 2016) coupled with innumerous international projects and investments in recent years, such as Living with a Star, NOAA/Space Weather Centre, International Space Environment Service, countless ground- and space-based solar observatories etc.

The CME type which has received the most attention by space physicists, is Earth-directed CMEs (commonly known as halo CMEs). These CMEs interact with the geomagnetic field, causing transient disturbances in the ionosphere which may affect satellite function (communication and navigation disruptions, increased orbital drag and damaged electronics) or even ground-based radio communication in the case of a particularly severe event (Schrijver et al., 2015). Consequently, the very first Pathway recommendation of the current COSPAR roadmap "...to advance the scientific understanding of the Sun-Earth connections leading to space weather" (Schrijver et al., 2015) is to advance current space weather research so that to achieve the capability to issue > 12 hour advance warnings of incoming CMEs which lies at risk to induce geomagnetic storms.

Forecasting the arrival time of a propagating CME requires information on a number of solar wind parameters, such as bulk speed, plasma pressure, density, temperature and magnetic field, at several points in the direction of propagation of the CME, as well as geometrical constraints of the erupting CME. Since there is only a limited number of operational spacecraft in the heliosphere which can provide in-situ measurements of the solar wind, in particular beyond the L1 point, it is absolutely necessary to make well-grounded numerical approximations and construct models which can simulate the solar wind with sufficient accuracy. However, the numerical prediction of CME parameters requires sophisticated magnetohydrodynamic (MHD) simulations and the predicted ar- rival time error is commonly on the order of 12-24 hours (Tucker-Hood et al., 2015). Testing the performance of solar wind models by comparing model output with in-situ measurements, and de- termining the main error sources which could be responsible for potential discrepancies is therefore of great importance, and the main objective of this comparative case study.

This case study concerns four CME events, their complex structure as is detected by in-situ

measurements at Earth and the geomagnetic disturbances they produced. Using existing simulation

runs of a coupled coronal-heliosphere model, the Wang–Sheeley–Arge (WSA)/Enlil+Cone model

installed at the Community Coordinated Modeling Center (CCMC), and consistent data sets from

the OMNI 2 database, a quantitative and qualitative comparison is made between data and model ouput. The quantitative comparison concerns the difference between the predicted and actual shock arrival time (SAT) and the amplitude of the corresponding shock, while the qualitative analysis covers in-situ signatures of the plasmoid’s internal structure, how well this is reproduced by the model and possible, extraneous interactions. The analysis aims to determine the main error sources of the prediction, as well as identifying future steps that should be taken to solve any systematical deviations between data and models.

Previous studies on the matter of ICME arrival time error and overall agreement between the WSA-Enlil+Cone model and data, mainly consists of statistical surveys (Mays et al., 2015a; Lee et al., 2015, 2013; Vršnak et al., 2014; Jang et al., 2014; Taktakishvili et al., 2011, 2010). These, so called, "ensemble" studies have considered the dependency between the SAT error and the choice of input parameters for ICMEs with well-defined geometrical parameters. Only one previous refereed article, however, has conducted a case study of a particularly complex event (Mays et al., 2015b), which yielded a false positive alert. This paper aims to make a contribution to the latter kind of study.

2 Review

2.1 CMEs 2.1.1 Definition

The prefix "coronal" in coronal mass ejections comes from the fact that CMEs were first observed with the use of white-light coronagraphs, which visualizes the Sun’s corona by obscuring the solar disk. It is suspected that CMEs does involve and later eject plasma and magnetic field from the photosphere and chromosphere, but the majority of the expelled mass is (indeed) expected to have its origin in the lower corona (Webb and Howard, 2012). The observational definition of CMEs is based on the following three properties: it appears as a sudden, well-defined brightening above the solar limb, it is accompanied by additional, bright material that moves outward in the coronagraph’s field of view and it occurs on a time scale of a few minutes to several hours (Webb and Hundhausen, 1987). The characterizing properties that this definition leaves out, such as CME speed, angular width and all the way to their initiation mechanism and propagation, are described in the rest of this section. A statistical analysis (Howard et al., 1985) have shown that CMEs appear to have a multitude of different structural profiles as seen from Earth. However, in recent years the common belief seem to be that this apparent diversity is largely due to projection effects (Schwenn, 2006).

The general perception today is that CMEs, at large, tend to possess a three-part structure shortly after its eruption consisting of a bright outer loop, a dark cavity and a bright kernel somewhat resembling a light bulb (see Figure 1). Physically, this structure has been interpreted as heavily compressed plasma pushed ahead of a visually dark, magnetic flux rope followed by a filament (Illing and Hundhausen, 1985).

2.1.2 Initiation

Many questions remain regarding the cause and physical processes involved in the onset of CMEs,

and it is still a matter of heavy debate. In its simplest form, its cause can be attributed some form

of instability in the coronal magnetic field. The eruption of a CME is therefore a means for the

Figure 1: The exceptionally clear three-part structure of a CME erupting on 2000-02-27, as imaged by the C3/LASCO/SOHO coronagraph (Solar and Heliospheric Observatory, 2000).

global magnetic field of the Sun to free itself of excess energy, which is likely to build-up during the rising phase of the solar activity cycle (Low, 1996). In most cases, it is believed to take the form of an elevated, stressed, local magnetic field structure, resembling a helix and befittingly called a flux rope (Webb and Howard, 2012). Plasma flows from the flux rope’s foot points in the Sun’s photosphere, and is entrained within the flux rope. At some point it has harbored enough energy to become unstable, erupt and escape the Sun, producing a CME as seen in the coronagraphs.

However, this description is far from certain, and does not describe how and where factors like magnetic reconnection and MHD instability come into play, whether the process occurs on a global or local scale, and the details regarding the evolution of its magnetic topology (Lebedev, 2007). The eruption of CMEs and the preceding stage leading up to the eruption, has since long been linked to numerous diverse phenomenon of solar origin, which potentially could be used as a prediction tool of an impending CME. A number of promising precursors, as they are called, include flares, filament eruptions, coronal dimming, sigmoids, Moreton and EIT waves as well as submillimeter radio bursts (Schwenn, 2006). The case for solar flares have perhaps been the most disputed one through the decades. It is now standard belief, however, that the simultaneous eruption of a solar flare and a CME does not necessarily reflect any connection between the two different processes, since many CMEs have been observed to occur far from active sunspot regions (Howard and Tappin, 2008). It is rather two different manifestations caused by the same general cause, i.e. an excess of magnetic energy (Kahler, 1992).

2.1.3 Kinematic properties

Hitherto, the apparent speed of the leading front of CMEs span the entire spectrum between 21-2625

km/s, when estimated close to the solar limb (SoHO LASCO CME Catalog, 2016). The propagation

speed that is measured by the coronagraph, however, is merely a projection of the speed along the

Sun-Earth line (which from now on will be denoted as the radial speed). Fortunately enough, a linear, empirical relationship seems to hold between the lateral expansion speed, for which a unique determination can be made for any CME, and the radial speed (Schwenn et al., 2005). This property of coherence, often called "self-similarity", also seem to hold at large well beyond the field of view of any coronagraph, which is something that will be elaborated upon in greater detail in Section 5. This self-similarity enables an estimation of the radial speed of virtually any CME, and thus appoints the speed as one of the most practical CME properties, for which relationships can be established regarding the further CME propagation and geoeffectiveness.

The uncertainties linked to the determination of the radial speed entail even greater problems for the derivation of CME mass and energy via coronagraph observations. For instance, it is as- sumed that the entire body of mass lies in the sky plane (Webb and Howard, 2012), which is to be considered an ill-defined assumption due to the earlier mentioned projection effects. Since the appropriate equations which determine the density of the object have an explicit direction term (Webb and Howard, 2012), it is clear that the mass derivations are associated with large uncer- tainties. Nonetheless, an independent mass verification can be performed through radio (Ramesh et al., 2003), X-ray (Rust and Hildner, 1976) and EUV light (Aschwanden et al., 2009) observa- tions. Also, such estimates have been cross-checked via the stereoscopic view of the two STEREO satellites (DeForest et al.), enabling a higher level of accuracy. The CME energy is a matter of vital importance for models concerning the eruption mechanism of CMEs and their geoeffectiveness.

However, since this parameter is linked with perhaps the greatest uncertainties of all, every estab- lished relationship involving this parameter has been very uncertain and it has been impossible to draw any firm conclusions.

2.1.4 Halo CMEs

The outward flow of plasma from a specific kind of CMEs seem to encompass the whole occulted

solar disk in the coronagraph field of view, giving the impression of a global eruption over the entire

solar disk (Webb and Howard, 2012). This appearance, however, is nothing more than a projection

effect in contrast to an actual physical phenomenon. This is realized by imagining the case of a

typical CME with a large component of its velocity along the Sun-Earth line. The outward plasma

motion of such an event would appear as an expanding circular brightening, a halo, on the corona-

graph. Thus we can conclude that these events are simply Earth-directed CMEs, which are called

halo CMEs for the above described reason (Howard et al., 1982). CMEs with an angular width

between 120-360

, which are partly directed towards Earth but not enough to be classified as a

halo CME, are referred to as partial halos (Yashiro et al., 2005). One danger with the classification

scheme based on angular width, however, is the fact that a CME directed in the complete opposite

direction to Earth (i.e. backsided) would also appear Earth-directed (frontsided) in the corona-

graph’s field of view (Webb and Howard, 2012). Therefore the localization of a possible source

region through another means of observation, such as in EUV, is necessary to rule out the possi-

bility of a false alarm. When no active regions or other possible precursors can be identified, the

stereoscopic capabilities of the STEREO satellites is of paramount importance in order to deduce

if the considered event is actually front- or backsided. Before the age of space-based coronagraphs

situated outside the Earth’s Lagrange point 1, situations such as the one described would be irre-

vocably associated with mistakes. The central position angle of frontsided halo CMEs is usually

located very close to the heliographic equator. The 10 % that have been detected closer to the so-

lar limb must be considerably more energetic to produce a halo CME event (Gopalswamy, 2010).

Their likely position at the solar meridian enables for the possibly associated precursor, such as a sunspot group, to be studied in greater detail (Webb and Howard, 2012). However, due to the infa- mous projection effects, properties such as the actual angular width is impossible to discern without the aid from stereoscopic observatories (Gopalswamy et al., 2015b). Since the angular width of halo CMEs is difficult to determine, it is the speed that serves as halo CMEs main characteriza- tion tool. Statistical studies have shown that halo CMEs tend to possess a higher sky-plane speed and thus be more energetic than their standard counterparts (Gopalswamy et al., 2010). This fact, however, is unlikely to mirror an actual speed difference, it is rather a matter of missed detections due to projection effects (Webb and Howard, 2012). What is definite however, is that the majority of the CMEs that have been deemed responsible for the most intense geomagnetic storm and solar energetic particle (SEP) events that have occurred in modern time, are in fact halo CMEs (Webb and Howard, 2012). This is surely strongly connected to their source region, as halo CMEs with a central position angle close to the meridian have been observed to be particularly geoeffective.

2.2 ICMEs

Historically, it has been custom to denote the propagating CME by interplanetary coronal mass ejection (ICME), as a measure by which to distinguish between the actual CME and CME-driven effects (Zhao and Webb, 2003). ICMEs have proven to be the main driver behind SEP events (Cane and Lario, 2006), intense geomagnetic storms (Zhang et al., 2007) and a regulator of the galactic cosmic ray intensity on Earth (Cane, 2000). Moving at a velocity greater than the magnetosonic speed of the solar plasma, they manage to drive collisionless shock waves as far as beyond Mars orbit (Zurbuchen and Richardson, 2006). The propagation causes severe compression, deflection and heating of the solar wind, notably causing a disruption in the typical Parker spiral topology (Schwenn, 2006).

2.2.1 The solar wind

In order to be able to recognize the arrival of an ICME at Earth, it is important to understand the physical processes that occur when an ICME propagates through the interplanetary medium.

Therefore, we make a brief description of the basic properties of the solar wind in the following paragraph.

The solar wind is a continuous flow of plasma from the Sun, which is the result of the ongoing

expansion of the solar corona (Gosling, 2009). It reaches far beyond the orbit of dwarf planet Pluto,

as it decreases in both density, pressure and temperature as a function of the radial distance from the

Sun. Due to the high, electric conductivity of the plasma the coronal magnetic field is swept along

with the solar wind, while its footpoints remain anchored at the Sun. Since the Sun rotates about its

own axis the solar wind’s magnetic field is dragged along, creating a magnetic topology identical to

an Archimedean spiral (known as the Parker spiral) (UCR Center for Visual Computing, n.d.). The

solar wind can, in turn, be divided into two subgroups depending on its radial speed. There is the

slow, dense solar wind, moving at an approximate speed of 350 km/s, and the fast, attenauted solar

wind, maintaining a speed of 750 km/s on average. The fast solar wind is thought to originate from

large-scale coronal holes, which have their magnetic field lines opened up toward interplanetary

space (Koskinen, 2011).

Figure 2: Schematic illustrating the formation of a CIR (Nikbakhsh, 2014).

Due to the spiral pattern of the solar wind, the solar wind speed varies in the same direction as a function of time. In such a situation, the fast solar wind catches up with the slow solar wind and an interface between the two solar wind types is formed, which is called a corotating interaction region (CIR) (Tsurutani et al., 2006). CIRs are characterized by a compressed plasma region bounded by two boundary layers, where the speed, density, temperature and magnetic field strength of the plasma experiences a discontinuous (or rather very abrupt) change (see Figure 2).

When the solar wind reaches 1 AU it encounters the geomagnetic field surrounding Earth. Since the magnetic field lines are "frozen-in" within the plasma of the solar wind, it will be deflected by the Earth’s magnetic field, creating a magnetopause surrounding the Earth’s magnetosphere (Tidman, 1967), as seen in Figure 3. Since the solar wind flow is supersonic (Richardson, 2010), a bow shock will form in front of the magnetosphere, where the solar wind is decelerated to subsonic values.

Downstream of the bow shock, in the magnetosheath, the speed of the plasma is decreased and some of the excess kinetic energy is converted into thermal energy (Argo et al., 1967). One might assume that due to the presence of the magnetopause, the solar wind could not affect the Earth in any manner. However, a process known as "reconnection" can act within the magnetopause (see Figure 3), causing the energetic solar wind to couple into the geomagnetic field (Tanskanen et al., 2012). In some cases, the charged particles that substitute the solar wind plasma may reach into the Earth’s thermosphere, creating polar lights. This is the particularly the case for ICMEs sparking intense geomagnetic storms (Guarnieri et al., 2006).

2.2.2 IP shocks

The bow shock created as a result of the interaction between the solar wind and the quasi-stationary

magnetosphere, or the interaction between the fast and the slow solar wind at a CIR, are both exam-

Figure 3: Schematic showing the different regions of the geomagnetic field (Merck, 2014). The yellow arrows signify the flow direction of the solar wind.

ples of interplanetary (IP) shocks. An IP shock is primarily defined as a (near) discontinuous change between two adjacent plasmas with a relative difference in speed, causing a similarly dramatic change in other plasma properties (such as density, temperature and magnetic field) (Nikbakhsh, 2014). It is required that the fastest plasma exceeds the magnetosonic speed of the medium in order to create a shock wave (Richardson, 2010).

IP shocks can be classified as a combination of forward/reverse and fast/slow. A forward (re- verse) IP shock is defined as a shock moving away (toward) the Sun in the solar wind reference frame, seen as an increase in velocity and an increase (a decrease) in density and temperature as detected by the observer (Andrioli et al., 2007). In the Earth’s frame of reference, both types of shocks will be seen as moving away from the Sun (Echer et al., 2003). Sometimes, it might be more intuitive to describe a forward IP shock as the leading edge and a reverse IP shock as the trail- ing edge of the same event. This is exactly the case for CIRs, where the boundary layers confining the compressed CIR are in fact detected as a forward-reverse shock pair (Howard, 2011a).

In the case of an regular, unperturbed ICME, the leading front creates a forward shock but not necessarily a distinguishable reverse shock, as there is nothing that can compress the ICME from behind. This is not necessarily the case if it could be affected by forthcoming CIRs or other ICMEs, which is a possibility that should not be neglected, as we’ll soon shall see. The classification

"fast/slow" refers to the behavior of the magnetic field at the shock, if the magnetic field strength increases (decreases) it is classified as a fast (slow) shock (WIND Magnetic Field Investigation, 2001; Echer et al., 2003).

The most commonly observed shocks at Earth are fast forward shocks, and thus most geomag-

netic storm have been attributed to CMEs rather than CIRs. This is due to the fact that most CIR

shocks, with their characteristic fast forward-reverse pair, seem to develop at a distance beyond 2-3

AU, while ICME shocks can be quite faster and thus form closer to the Sun (Richardson, 2010).

Slow IP shocks of either type are considerably more rare (WIND Magnetic Field Investigation, 2001). In fact, it has been noted that the number of detected reverse IP shocks is higher during the declining phase of the solar activity cycle. This peculiarity can be attributed to the fact that the polar coronal holes are more well-developed and the speed difference between the fast and the slow solar wind increases, along with the fact that the tilt of the Sun’s magnetic axis tend to increase, at the solar minimum, creating more CIRs in unison (Richardson, 2010).

2.2.3 In-situ signatures

There is undisputable need for a means by which the arrival of an ICME can be detected, both from a stereoscopic and Earth-bound (L1) perspective, so that precautions can be taken to minimize and assess the potential damage of an ICME impact. Space-based satellites such as the Advanced Composition Explorer (ACE) and the twin STEREO satellites fulfill this need by continuously taking in-situ measurements of plasma properties and the IMF.

Studies conducted through the years have shown that the passage of an ICME can be detected by a vast number of so called "signatures", i.e. distinct property changes implying the presence of a structure separated from the regular solar wind (Zurbuchen and Richardson, 2006). These signa- tures can be loosely divided into a few categories, depending on whether they are associated with plasma dynamics, magnetic field, plasma composition, plasma waves and suprathermal particles (Zurbuchen and Richardson, 2006).

We refrain from delving any deeper into the three latter categories within the scope of this review, other than simply mentioning a few noteworthy examples of such signatures: suprathermal (Gosling et al., 1987) and bidirectional electrons (Palmer et al., 1978), cosmic ray depressions (Cane et al., 1994) and an overabundance of helium (Hirshberg et al., 1971). Instead, we shall focus on the two categories connected to plasma dynamics and magnetic topology for the remainder of this review (see Figure 17).

A noteworthy aspect of ICME signatures, is that there exists a great variation in the number of signatures that are detected for each individual case (Gosling, 1993). More often than not, only a few signatures of the ones listed in Figure 17 can be detected for each event (Richardson and Cane, 2010). Some signatures, like an anomalous decrease in the proton temperature, are observed for virtually every event while others (many belonging to the three latter categories) have only been observed in a few cases (Skoug et al., 1999). The possibility of CME-CIR and CME-CME interaction cause further difficulties.

2.2.4 Structure: Sheath

As mentioned in the earlier subsubsection, the velocity by which the leading edge of ICMEs prop-

agate through the interplanetary medium is commonly of a higher order than the surrounding solar

wind. For this reason, the leading edge is detected as a fast forward shock at Earth. Just behind

the leading edge, and beyond the fast forward shock, a region known as the sheath (or a number of

sheaths) begins (see Figure 5). This structure is characterized by a high density and an enhanced

magnetic field strength, but most importantly, an extremely high temperature and turbulent behavior

in the velocity and magnetic field components (Burlaga et al., 1981; Richardson and Cane, 2010).

Figure 4: Table summarizing the main in-situ signatures of ICMEs connected to magnetic field (B),

plasma dynamics (P), plasma composition (C), plasma waves (W) and suprathermal particles (S)

(Zurbuchen and Richardson, 2006).

This is due to the fact that this region, although not as extremely compressed as the leading shock front, is still highly compressed, containing hot turbulent plasma and a pile-up of magnetic field lines, which then flow outward on both sides of the main body of the CME. In fact, since the magnetic field is both enhanced and extremely variable within this region, it has been suspected to be a contributor or sole driver of a vast number of geomagnetic storms, in particular for ICMEs that lack clear in-situ signatures of a magnetic cloud structure (Tsurutani et al., 1988). Field line draping (McComas et al., 1988) along the sheath may also cause distortions of the Parker spiral IMF topology in the heliosphere and complicating the modeling of subsequent heliospheric structures.

Figure 5: Schematic illustrating the theoretical structure of an ICME (Zurbuchen and Richardson, 2006).

2.2.5 Structure: Magnetic cloud

The following region, the "cavity", contains the majority of the bulk mass (although declining in density), and often, the main magnetic structure of the initial CME (Burlaga et al., 1981). It is most easily detected via an enhanced alpha to proton ratio, which is beyond the scope of this review, or via a decreased proton temperature, which has been detectable in virtually all CME events so far.

Approximately 30-50 % of all detected ICMEs have revealed an instrinsic, magnetic structure of helical configuration, reminiscent of the flux rope of the associated CME (Burlaga et al., 1981;

Gosling, 1990; Cane et al., 1997). This subtype of ICMEs, known as magnetic clouds, are charac-

terized by a very high magnetic field strength, inherently low temperatures and a gradual, smooth

change in polarity of one or several magnetic field components, indicating the changing pitch angle of a loop (Bothmer and Schwenn, 1998). Also, the low magnetic field variance within the magnetic cloud differ significantly from the chaotic magnetic field component behavior in the preceding sheath (Enzl et al., 2012). It has been proposed that perhaps all ICMEs are magnetic clouds (Webb and Howard, 2012). If so, it has been speculated that ICMEs lacking the above in-situ signatures would simply arise from a glancing encounter, i.e. a case where merely the "leg" of the flux rope manages to pass through the satellite or that the structure of the flux rope is somehow obscured or even less structured than usual (Webb and Howard, 2012). There is also a possibility that a CME-CME interaction may distort the clean signature of a single flux rope or that the magnetic structure of the ICME was more complicated to begin with (Liu et al., 2015). The disappearance of well-developed filaments on the solar disk have been associated with the few detection of magnetic clouds, supporting the hypothesis that the magnetic field structure within magnetic clouds is indeed the actual flux rope of the original CME (Webb and Howard, 2012; Marubashi, 1986). More so, evidence has been presented from a multitude of studies that the detected flux rope seems to retain the same polarity and orientation as the original CME flux rope (Marubashi et al., 2015; Yurchyshyn et al., 2007).

2.2.6 Structure: T

depression

One of the most well-used ICME signatures, consisting of an atypical decrease in proton temper- ature (designated "T

depression") during the passage of the ICME cavity (most likely a magnetic cloud), is detected for the absolute majority of all ICME passages (Richardson and Cane, 1995).

T

depression can be made quantified with the use of the following condition, T

T

≤ 0.5 (1)

where T

is the expected proton temperature, as calculated by the following empirical relationship between T

and the speed of the ambient solar wind (Lopez and Freeman, 1986; Richardson and Cane, 1995),

T

=

( (0.0106 v − 0.278)

/R, if v < 500 km/s

(0.77 v − 265)/R, otherwise. (2)

The leading edge of the magnetic cloud, or cavity, is then discerned as the point at which this condition is fulfilled. The end of this stretch is interpreted as the trailing edge of this region, which can be confirmed by the appearance of a more variable magnetic field.

3 Data and Method

This case study concerns the agreement between data and model at the Sun-Earth Lagrange point

L1 for three geomagnetic disturbances and its connection to four individual halo, or partial halo

CMEs. The initial idea was to investigate the agreement for Earth-directed ICMEs at three different

locations (ACE, Stereo A and Stereo B) at a time when the angular separation between Stereo A

and B was 90

≤ θ ≤ 180

. However, it was found that runs from the WSA-Enlil+Cone model

were only available at a time when the angular separation had surpassed 180

, and so there was

no choice left but to select ICMEs for those large angles. We use a data set of high-quality model

output from the WSA-Enlil+Cone (courtesy of Prof. Hermann Opgenoorth) for the time frame 2012-01-20 until 2012-01-30 (Odstrcil, 2016). Three Earth-directed ICMEs, with two resulting IP shocks and geomagnetic disturbances, are picked from this data sample. A final event is chosen for comparison, namely the so called "St Patrick’s storm" of 2015 March 17, this time however without the same range of data products. Unfortunately it was found that the said model output did not contain information on the input values of each run, such as initial speed and half-width of the CME.

Such information had to be collected from the second best option, the Space Weather Database of Notifications, Knowledge, Information (DoNKI). However, the runs listed for the appropriate events in said database was likely of an earlier version, see Section 5 for a further discussion on this matter.

In-situ measurements of the solar wind are collected from the OMNI 2 data set, which contains online-accessible, multi-source data of a wide range of plasma parameters and the IMF at a hourly resolution (GSFC Space Physics Data Facility, 2016a,b). The OMNI 2 data set picks the best data out of all available spacecraft at L1 for each data point, keeping data gaps to a minimum. For the studied events, the Comprehensive Solar Wind Laboratory for Long-Term Solar Wind Measure- ments (Wind) and The Advanced Composition Explorer (ACE) are the main contributors for all collected in-situ data (GSFC Space Physics Data Facility, 2016c,d).

Since the range of the model output is quite limited, the following parameters were chosen from the relatively large OMNI 2 data set: the (proton) number density N (unit: cm

), (the radial component of) the proton speed v (unit: km/s) (the radial component of )the proton temperature T (unit: kK), the normal RTN-component of the IMF B

(unit: nT) and the magnetic field strength

|B| (unit: nT).

The southward component (B

) of the IMF in the geocentric GSE or GSM coordinate system is equivalent to the normal component (B

) in the space-craft oriented RTN coordinate system (Alexandersson, 2011). These different designations are used interchangeably within the context of this study.

The circumstances concerning each event is examined, as well as any potential mutual inter- action. The full course of action, i.e. the preceding state, eruption and the perceptible part of the subsequent propagation, is examined via a set of online resources (Spaceweather.com, 2016; So- larMonitor, 2016; HelioViewer, 2016; SolarSoft, 2016). Any surrounding circumstances that might directly, or indirectly, have affected the events in an atypical manner, and as such could affect the agreement between data and model, are carefully noted.

The model output is collected from HelioViewer (2016), and is loaded into MATLAB (2013)

along with the corresponding data. The prediction and data is plotted together, and miscellaneous

calculations are performed for the qualitative and quantitative analysis of the events. The shock

arrival time (SAT) for both model and data were inferred by the use of the "diff" MATLAB function,

which enabled for a localization of the fast forward shock of each detected IP shock. The SAT

error was then computed as the difference between the predicted and the actual SAT. Using the

empirical formula concerning the conditions for the Tp depression that is commonly observed in

ICME events, regions of possible Tp depression could be plotted. For the quantitative analysis, the

input speed of each event was used to investigate any potential proportionality with the SAT delay,

or the total travel time.

3.1 The WSA-ENLIL+Cone Model

The simulated data sets utilized in this report is the product of the WSA-ENLIL+Cone model,

implemented at NOAA Space Weather Prediction Centre and available upon request from the Com-

munity Coordinated Modeling Center (Pizzo et al., 2011). The model offers real-time modeling of

ICME propagation in a solar wind environment based on actual high/low speed solar wind streams,

the realistic time-evolution of magnetohydrodynamic processes in three dimensions and does auto-

matic calculations of the estimated SAT. The model also takes into account the interaction between

CMEs propagating in the same domain of the heliosphere and provides model output at the real-

time position of all planets and all major satellites within the inner part of the solar system. Such

a range of data output alternatives and the level of agreement with actual physical processes and

observed phenomenon is unprecedented, when judged against comparable models currently in-

stalled at CCMC. The WSA-ENLIL+Cone model is a coupled coronal-heliosphere model, where

the Wang–Sheeley–Arge (WSA) coronal model and the geometric Cone model provides the bound-

ary conditions for the heliospheric ENLIL model that describes the majority of the ICME propaga-

tion. The WSA model is generated from daily synoptic magnetograms and generates a prediction

of the background solar wind, namely the IMF and speed profile at 21.5 R, which then is used

as input for the ENLIL model which, in turn, is used to reproduce the entire heliospheric view

of the ambient solar wind. Alternative coronal models installed at CCMC, such as Magnetohy-

drodynamic Algorithm outside a Sphere (MAS) or heliospheric tomography from interplanetary

scintillation (IPS), have and are being been coupled with the ENLIL model. However, as the avail-

able predictions were exclusively ENLIL-WSA+Cone models, we will refrain from delving any

deeper into the named alternatives. The Cone model provides an approximation of the actual, ge-

ometric properties of the ICME. Making use of the self-similarity property for CMEs, it assumes

(as its name suggests) a cone-shaped ICME with isotropic expansion, radial propagation, a constant

CME cone angular width, The input parameters to the Cone Model are determined by a geometric

triangulation algorithm, using real coronagraph observations from multiple vantage points. The

output includes an estimate of the expected arrival time (at the inner boundary), cone CME latitude,

cone longitude, CME width and the space speed in three dimensions. All of the mentioned output

parameters are determined at the 21.5 R boundary. In the simplest case, the output information is

launched into ENLIL as series of homogeneous wedges, making up a spherical plasmoid with all

the kinematic and geometric properties given by the Cone model. The Cone model input parameters

in the case of halo CMEs are deduced with the use of beacon-level data from the COR2/SECCHI

onboard the twin Stereo satellites. The COR2 data can be turned into running difference images,

which in turn are turned into time-elongation maps, from which the speed can be estimated and

then cross-checked via the Liu triangulation method. The cone attributes can also be determined

via said triangulation method, but this time with the use of the Stereoscopic CME Analysis Tool

(StereoCAT). This procedure, however, requires data from both spacecrafts at a comparable time,

and a moderate separation angle. ENLIL makes up the heliospheric part of the coupled model, and

enables the tracking of an ICME in three dimensions within a radial distance of 21.5 R out to

10 AU. The ICME is interpreted as a disturbance of the ambient solar wind, which is enabled to

propagate through the heliosphere. Free parameters include the cavity ratio, defined as the ratio of

the radial CME cavity width to the CME width, with the default being no cavity. The CME cloud

density is a another free parameter.

3.2 Geometric triangulation

As was first proposed by (Sheeley et al., 1999) it is possible to derive the distance, velocity and direction along which an ICME moves through the heliosphere from time-elongation diagrams (so called "J-maps", see Figure 7). At the time, Sheeley et al. was restricted to the less advantageous, single viewpoint offered by LASCO/SOHO. Since then, a number of stereoscopic versions of said technique have been suggested, the first one being the Geometric Triangulation technique developed by (Liu et al., 2010). With the use of coronagraphs (COR2) and heliospheric imagers (HI1 and HI2) onboard the twin Stereo satellites, it was proposed that it would be possible to both track and infer kinematical properties of a white-light feature moving between said spacecrafts. The model makes the assumption that the same point-like feature is identified and tracked by the two spacecrafts, and that the full motion is confined to the ecliptic plane. The validity of said assumptions is discussed in Section 5. Running-difference images, which depict instantaneous changes in the original white- light images, are stacked together for the full time period of propagation in the heliosphere. The elongation (α

and α

in Figure 6), as seen from each respective spacecraft, is then plotted against time to produce a time-elongation map. The described technique assumes that the angles β

and β

(see Figure 6) are held constant during the entire movement, motivated by the self-similarity claim. ICMEs appear as slightly bent tracks in these J-maps, and extend to large elongation angles (> 60

for events crossing the Earth’s orbit). By comparing the slope and angle β of a single track from each of the two satellites, its velocity and direction of propagation along the ecliptic plane can be determined. Auxiliary versions of such J-maps functions are used for the computation of input values to the Cone model.

Figure 6: Schematic of the ecliptic plane, including the geometric properties that are involved in the

Liu triangulation algorithm (Liu et al., 2010). The point feature that is tracked by the twin Stereo

satellites is denoted as P. Note that the schematic does not depict the exact geometric situation for

the studied events in this paper, as implied by Figure 8.

Figure 7: After-processed J-map (Odstrcil, 2016) spanning the time of propagation of the 18, 19 and 23 January events. The yellow arrow indicate the appearance of the January 19 event (possibly intertwined with the January 18 event), while the red arrow point out the track from the January 23 event.

Figure 8: Positions of the Stereo A and Stereo B satellites at a) 2012-01-19 15:10 UT and b) 2015- 03-15 02:00 UT (STEREO Science Center, 2016), as seen from above/below the equatorial plane.

Both axes makes use of the unit AU (astronomical unit = distance between the Sun and the Earth).

4 Results

A summary of the input parameters, predicted and actual SAT, travel time etc. for each of the studied events is found in Table 1. The full set of graphs that depicts the model output and the in-situ data for each of the chosen parameters, together with the highligthed SAT and possible T

depression, are found in Figure 9–11 and Figure 13–16. Figure 12 depicts the difference between T and the expected temperature values, T

. Figure 18 consists of a proportionality analysis between the input speed and the SAT delay, as well as the total travel time. The full code can be accessed online (Werner, 2016). The following subsections deals with the observations, and related circumstances, of each studied CME event and subsequent IP shock.

Table 1: Summary of the input and output parameters for the selected events.

Y

2012 2015

E

Jan 18 13:25 UT Jan 19 15:10 UT Jan 23 04:00 UT March 15 02:00 UT

I

[

/

] 520 1020 2211 750

H

-

[

] 32 69 62 45

P

SAT Jan 22 03 UT Jan 22 03 UT Jan 24 14 UT March 17 15 UT

R

SAT - Jan 22 06 UT Jan 24 14 UT March 17 04 UT

Figure 9: The radial speed and the number density profile of the 2012 January 22 and 24 IP shock

events. The measured number density profile is shown in red, while the predicted profile is shown

in orange. Likewise, the measured radial speed profile is shown in blue, while the predicted profile

is shown in turquoise. The shaded regions represents a time period of temperature depression.

Figure 10: The magnetic field strength and temperature profile of the 2012 January 22 and 24 IP shock events. The measured temperature profile is shown in red, while the predicted profile is shown in orange. Likewise, the measured magnetic field strength is shown in blue, while the predicted profile is shown in turquoise. The shaded regions represents a time period of temperature depression.

4.1 The 2012 January 18 event

A partial halo CME event occurred on 2012-01-18 at 13:25 UT, which was detected by a total of four space telescope instruments: COR2/SECCHI/STEREO A, COR2/SECCHI/STEREO B, C2/LASCO/SOHO and C3/LASCO/SOHO. The partial halo was estimated to a half-width of 32

and an initial velocity of 520 km/s. It was expected to reach the 21.5 R boundary on 2012- 01-18 20:20 UT. No obvious source region or associated solar events could be determined. The archives do not indicate the presence of an associated IP shock or geomagnetic storm. However, it was expected to merge with the following halo CME which itself would cause a disturbance. The details regarding this particular event’s contribution to the IP shock caused by the following halo CME event is discussed in Section 5.

4.2 The 2012 January 19 event

Following the eruption of the partial halo described above a halo CME erupted on the 2012-01-

19 15:10, being suspected to have arisen from a source region identical to that of AR11402 (lo-

cated at N32E27). The CME was detected by C3/LASCO/SOHO, COR2/SECCHI/STEREO A and

COR2/SECCHI/STEREO B, and was estimated to have a half-width of 69.0

and a speed of 1020.0

km/s. Passage of the 21.5 R boundary was predicted to occur at 2012-01-19 18:34 UT. The halo

CME event was somewhat associated with a M3.2 GOES class solar flare at AR11402. However, as

the flare had its peak X-ray emission at 16:05 UT its connection to the halo CME can be considered

open for discussion. The first WSA-ENLIL+Cone run predicted an impact on 2012-01-21 22:29

Figure 11: The profile of the magnetic field strength and the southward magnetic field component (B

= B

) from the 2012 January 22 and 24 IP shock events. The measured southward magnetic field component is shown in red, while the predicted profile is shown in orange. Likewise, the measured magnetic field strength is shown in blue, while the predicted profile is shown in turquoise.

The shaded regions represents a time period of temperature depression.

However, a later simulation, this time including the Jan 18 event which was expected to in- teract and then merge with said Jan 19 event, predicted another SAT for 2012-01-22 00:52 UT (±7 hours), with a 9.0 (±8) hour disturbance. It was also expected to hit Mars at 2012-01-24 06:04 UT and Spitzer at 2012-01-22 01:42 UT. It was anticipated that NASA’s Curiosity rover would be able to detect said disturbance. The actual IP shock was detected on 2012-01-22 06:17 UT, 3 hours later than anticipated. Its impact caused a compression of the geomagnetic field, and reportedly managed to interfere with satellites in geosynchronous orbit (GEO). Amateur radio en- thusiasts reported a vertical shift in the ionization layers in the Earth’s ionosphere, with a possible, negative effect on high-frequency radio communication during the time of geomagnetic disturbance (Spaceweather.com, 2012). Unusually strong ground currents was also reported in Lofoten, Norway (Spaceweather.com, 2012).

4.3 The 2012 January 22 IP shock

Relatively slow-speed solar wind (approx. 325 km/s), precedes the IP shock arrival on 2012-01-22.

The predicted SAT, indicated by the red dotted line in Figure 9, appear at 03 UT while the first

instantaneous speed change in reality occurs at 06 UT (as marked with the leftmost blue dotted

line). The proton speed reaches a maximum value at approx. 11 UT. It seems like the predication

managed to overestimate the solar wind speed by approx. 30 km/s and, perhaps consequently, the

amplitude of the peak on Jan 22nd by a figure of 70 km/s. The underestimated rise in speed at

approx. 02 UT on 2012-01-23 is the likely cause for the leveled appearance of the following, quasi-

monotonous speed reduction (as predicted by the model), lasting until the arrival of a fast forward

IP shock at approximately 14 UT on 2012-01-24.

Figure 12: The temperature profile of all selected events. Marked in blue is the temperature profile inferred from in-situ measurements and marked in green is the expected temperature profile as in- ferred from the basic relationship described in Equation 1. The shaded regions satisfy the condition written out in Equation 2.

Concurrent with the observed speed change at 06 UT on 2012-01-22, a steep > 40 cm

rise in the number density is detected (see Figure 9). The prediction underestimates the number density of the preceding solar wind by approx. 7 cm

as well as the amplitude of the fast forward shock at 06 UT by approx. 35 cm

. Coincident with the speed maximum at 11 UT is a density minimum, which is followed by three subsequent steep density peaks, the last one occurring at approx. 02 UT on 2012-01-23, neither of which was predicted by the model. Thereafter comes a more or less uniform low-density valley, simultaneous with the monotonous speed reduction as described above.

However, the model seems to have expected a density profile that is far lower than what what was actually observed. In fact, the model predicts a minimum density value > 1 cm

during said time period.

The temperature profile of the 2012-01-22 event (see Figure 10) seem to exhibit two, distinctive

peaks at approximately 11 and 21 UT on 2012-01-22. Followed by the second temperature peak

is a period of abnormally low temperatures, i.e. a proton temperature depression (as indicated by

the grey shaded area). The magnetic field strength, which is seen in the same figure, displays a

continuous rise between Jan 21-22 (not unlike the density increase during the same time period),

an oscillating behavior just past the fast forward IP shock peak, followed by a drop in magnitude

and then a local maximum in the late hours of January 22nd. Ensuing said local maximum is a quasi-monotonous decline in magnitude until approx. 19 UT on 2012-01-23, where the magnitude experiences a gradual increase until the SAT of the halo CME erupting on the 2012-01-23. Interest- ingly enough, the prediction appears to underestimate the amplitude of the magnetic field strength at 06 UT on 2012-01-22 by a factor of 1/2. The global decrease after the IP shock arrival is actually quite well represented by the prediction, if not for the fact that it is simply a scaled down version of the actual event. This might explain that the predicted dip in magnitude on the 2012-01-23 is approximately 18 hours too early. What is not expected by the model is the oscillating behaviour on 2012-01-22 and the moderate rise initiated at 19 UT on 2012-01-23.

The simulated profile of the southward magnetic field (see Figure 11) is underestimated by a factor of 10 and fail to depict the numerous spikes in magnitude that is demonstrated in the data.

The rise in the magnetic field strength between Jan 21-22, preceding the IP shock at 06 UT on 2012-01-22, seems to be largely due to an analogous rise in the B

component of the magnetic field. The southward component (B

) experiences three shifts in polarity after the arrival of the IP shock, after which it remains on the negative side of the diagram until approx. 10 UT on 2012- 01-23. However, the variance of the B

magnetic field component does not drop until 00 UT on 2012-01-23, lasting until the rise at 10 UT. Thereafter it showcases a pretty stable, low-variance behavior, if not for the sudden drop on 2012-01-24.

4.4 The 2012 January 23 event

04:00 UT at 2012-01-23 the eruption of a halo CME was detected simultaneously by C3/LASCO/SOHO

and COR2/SECCHI/STEREO B, with the suspected source region AR11401, which at the time was

located at the heliographic position N26W41. The CME was estimated to have a half-width of 62.0

and an initial speed of 2211.0 km/s. It was expected to reach the 21.5 R boundary on 2012-01-23

at 05:20 UT. Associated with the halo CME event was the eruption of a M8.7 GOES class solar

flare, with a peak time at (2012-01-23) 03:59 UT, and appearing from the adjoining active region

(AR)11402. The said active region was located at solar coordinates N33

W21

at the time of said

flare. The following SEP event, occurring between 04:49-05:30 UT on Jan 23rd, caused a solar radi-

ation storm of class S3. Such a radiation storm intensity is associated with potential moderate-level,

biological risks, i.e. an elevated radiation exposure for astronauts carrying out extravehicular activ-

ity (EVA) and passengers on high-latitude aircraft flights, and a slight effect on satellite operations

and HF radio propagation (NOAA, 2016). For this reason, Delta Air Lines reportedly had to reroute

their transpolar flights (Phys.org, 2012). The corresponding WSA-ENLIL+Cone run yielded a pre-

dicted SAT at Earth of 2012-01-24 at 14:18 UT, with a 7.7 hour disturbance. It was also predicted

to hit Mars at 2012-01-25 22:46 UT. The IP shock was first detected by ACE/MAG on 2012-01-24

at 14:30 UT. Interestingly enough, it was not detected by the Wind satellite. A moderate (class

G1) geomagnetic storm commenced at 2012-01-23 18:00 UT, reaching a Dst minimum of -73 nT

at 2012-01-25 11.00 UT. Although relatively moderate, the geomagnetic storm was strong enough

to be considered the largest one since the infamous "Halloween solar storm" of October 2003. A

more detailed analysis was made in 2013, which indicated the occurrence of another M-class solar

flare and a subsequent flux rope ejection, erupting at a comparable time and from the same active

as the main event, described above.

Figure 13: The radial speed and the number density profile of the 2015 March 17 IP shock event.

The measured number density profile is shown in red, while the predicted profile is shown in orange.

Likewise, the measured radial speed profile is shown in blue, while the predicted profile is shown in turquoise. The shaded regions represents a time period of temperature depression.

Joshi et al. (2013) argued that the intense SEP event and moderate geomagnetic disturbance that followed might be the result of an interaction between the two events, and thus most likely con- tributing to the event’s complex in-situ profile (see Section 5).

4.5 The 2012 January 24 IP shock

Even though the Jan 23rd event is notorious for being a particularly fast halo CME event, the model prediction manages to predict the correct hour of SAT, namely at 14 UT on 2012-01-24 (displayed by the second, red dotted line in Figure 9). This might be attributed to the fact that the model seems to have managed to estimate the speed of the pre-event solar wind stream quite well. However, the prediction does result in an overestimation of the amplitude of the fast forward IP shock by approximately 70 km/s. The predicted profile of the ICME front matches the data quite well, but the subsequent local maxima occurring at midday on Jan 25th is overestimated in amplitude by the model. The post-event speed value lies approx. 100 km/s above the corresponding speed of Jan 21st. The expected density profile of the Jan 23 event does not provide a good match to the data, as the number density instead seems to oscillate around the density value of Jan 21st. The temperature amplitude (see Figure 10) was also overestimated by the model, reaching a value of

>1600 kK instead of the approximate value of <8000 kK which is inferred by the data. The event never experiences a T

depression in accordance with the used definition. The data of the magnetic field strength during the event does not display the expected, sharp rise. Instead it seems to exhibit a two-step rise, with the second increase in the magnetic field strength possibly being associated with the second main peak in the temperature profile. The magnetic field components exhibit a considerably more noisy behavior when compared to the previous January event (see Figure 11).

The prediction, once again, fails to depict the correct magnetic field profile.

4.6 The 2015 March 15 event

On 2015-03-15 at 02:00 UT a halo CME event was detected by C2/LASCO/SOHO and C3/LASCO/SOHO from the source location S15

W24

on the solar disk, likely associated with AR12297. The CME was estimated to have a half-width of 45.0

and a speed of 750.0 km/s. It was expected to reach the 21.5 R boundary on 2015-03-15 at 06:45 UT. Associated with the halo CME event was the peak X-ray emission of a C9.1 GOES class solar flare, appearing from the same active region but at the position S22 W29. Also, a filament eruption and a bright post-flare arcade has been inferred from observations. The WSA-ENLIL+Cone model run predicted a glancing blow, with a predicted SAT at 11:39 UT on 2015-03-17 . The IP shock was detected by ACE/MAG on 2015-03-17 at 04:05 UT, five hours earlier than what was predicted. A two-step geomagnetic storm commenced at 2015-03-17 06:00 UT, reaching a global Dst minimum of -228 nT, which classifies it as a se- vere, class G4 storm, rendering it "the most severe geomagnetic storm of the current solar cycle"

(Gopalswamy et al., 2015a). This is to be considered exceptionally impressive, considering that only a glancing blow from the CME flank was predicted by the model. See NOAA (2016) for a table over the possible effects to power systems, spacecraft operations and other systems of such a geomagnetic disturbance.

The disturbance was likely intensified by the concurrent high-speed solar wind stream (max- imum velocity: 700 km/s), first detected by on ACE/SWEPAM on 2015-03-16, 00:00 UT, as the geomagnetic disturbance continued for more than 9 hours and aurora sightings were reported for six days subsequent to the initial IP shock. A recent study suggest that the 2015 March 15 event could be a CME-CME interaction event, as another CME, prior to the main one which erupted at 04:00 UT, can possibly be discerned in the C2/LASCO/SOHO footage (see Figure 14). However, since neither Stereo A nor Stereo B was functional at the time, it was not possible to infer any kinematical properties of said event.

Figure 14: C2/LASCO/SOHO footage from the eruption of the 2015 March 15 CME (SoHO LASCO

CME Catalog, 2016). Outlined in the two snapshots is the possible evolution of two separate CMEs.

Figure 15: The southward magnetic field component and temperature profile of the 2015 March 17 IP shock event. The measured temperature profile is shown in red, while the predicted profile is shown in orange. Likewise, the measured southward magnetic field component is shown in blue, while the predicted profile is shown in turquoise. The shaded regions represents a time period of temperature depression.

4.7 The 2015 March 17 IP shock

Preceding the fast forward IP shock arrival on 2015-03-17, the ACE satellite appears to be located in the midst of a low-speed solar wind stream (see Figure 13), reaching an approx. speed of 400 km/s just before the observed SAT at 04 UT on March 17th. The expected SAT (2015-03-17 15 UT) differs remarkably from what was inferred by observations. The simulated amplitude of the shock provides a better match to the data, although with a slight underestimation. The speed appears to rise quite a bit after the initial shock, reaching a local maximum of 600 km/s at approx. 12 UT on 2015-03-17. Thereafter, the model provides a generally good match of the speed profile.

Following the event, the speed rises to a global maximum of approximately 675 km/s, after which it settles at just below 600 km/s. The notable difference between the pre- and post-event speed likely indicates the involvement of a CIR in the ICME event, as is discussed in Section 5. The pre-event state of the IP shock is further characterized by a highly structured density profile (see Figure 13). The number density seems to oscillate around 20 cm

, but the oscillations grow in amplitude as the IP shock propagates toward Earth. The underestimated amplitude of the IP shock is probably largely due to the poorly simulated profile of the ambient solar wind preceding the event.

Immediately subsequent to the event, the number density drops down to a value slightly above that

of the ambient solar wind prior to the event, until approximately 17 UT on 2015-01-17, after which

a steep, monotonous decline occurs until the early hours of 2015-03-18. After the event, the density

flattens at approximately 4 cm

, further supporting the passage of a CIR.

Figure 16: The profile of the magnetic field strength and the southward magnetic field component from the 2015 March 17 IP shock event. The measured southward magnetic field component is shown in red, while the predicted profile is shown in orange. Likewise, the measured magnetic field strength is shown in blue, while the predicted profile is shown in turquoise. The shaded regions represents a time period of temperature depression.

The simulated temperature profile of the event (see Figure 15) is poorly matched by the data, since the observed amplitude is approximately 700 kK above that of the simulation. Also, the temperature spike is fairly short, ceasing at 15 UT on 2015-03-17, after which a quasi-depression (not quite fulfilling the empirical condition) in proton temperature occurs until 04 UT at 2015- 03-18. Afterwards, a rise in temperature occurs, settling at just below 200 kK after the event.

Once more, we have a case where the predicted magnetic field strength does not measure up to the observed amplitude (see Figure 16). In this case, the magnetic field strength displays, at first, a local maximum (connected to the IP shock) of 22 nT instead of the predicted 6 nT, and then continues to rise to a global maximum of >30 nT before declining. Also, the rise in amplitude of the magnetic field strength just preceding the IP shock is not grasped by the model.

The southward magnetic field component is the parameter that perhaps displays the most un-

expected behavior (see Figure 15 or Figure 16). Not only is the amplitude of the IP shock almost

10 times higher than what was expected (19 nT compared to 2 nT), it also exhibits a steep (11 nT)

second rise and subsequent dip which was not expected by the model. This dip, initiated at 12 UT

on 2015-01-17, remains in the negative part of the diagram (see Figure 15) until 00 UT on 2015-

03-18, reaching a global maximum at about -20 nT. After the dip, the magnetic field component

continues to fluctuate between 5 and -5 nT until the end of the simulation. There are a few time

gaps during which the proton temperature depression condition seems to be fulfilled, being best

satisfied between 14 UT on 2015-03-17 and 04 UT on 2015-03-18 (see Figure 12). As this time

period is almost perfectly coincident with the second dip in the B

component, it is reasonable to

assume a passage of a magnetic cloud within this time frame.

5 Discussion

5.1 2012 January 22

The time period immeadiately following the observed SAT indicated by the leftmost blue dotted line in Figure 9– 11 distinguishes itself from the mentioned IP shock front by a local maximum in speed and temperature, a local minimum in density, and a oscillation in both the magnetic field strength and its southward component. All of these signatures seem to suggest the presence of a sheath region. The time period immediately following this point until 02 UT on 2012-01-23, is characterized by several jumps in density, magnetic field strength, temperature, magnetic field strength and polarity. This region is more ambiguous, and could be interpreted as a prolonged sheath region of a single ICME; a front-sheath-front interaction;the unfinished merging of two ICME sheaths; or something entirely different.

The temperature depression occurring between 02 UT on 2012-01-23 until 14 UT on 2012-01- 24 (neglecting the occasional gaps) coincides with a declining number density and radial speed.

The variance of the magnetic field strength, and B

, seem to decrease during this time period, in particular until 10 UT on 2012-01-23. Most importantly, the southward magnetic field component appears to experience a gradual change in polarity until 10 UT on 2012-01-23. These features, in conjunction, seem to suggest the passage of a magnetic flux rope, i.e. a magnetic cloud.

Figure 17: Snapshots from the WSA-Enlil+Cone model simulation of the Jan 18 and Jan 19 ICMEs

as they propagate through the heliosphere. The view is centered at the ecliptic plane and the col-

orbar vizualizes the normalized solar wind density. The snapshot taken at 08 UT on 2012-01-20

(a) displays the distinct features of the two Earth-directed CMEs. The second snapshot (b) instead

dispays the expected merging of said plasmoids. As is clearly seen in the two images, these two

events were not expected to hit the twin Stereo satellites.

The time period between 10 UT, 2012-01-23 and 14 UT, 2012-01-24 displays a more atypical behavior, including a sudden drop in the B

component, a parabolic behavior by the magnetic field strength and a few gaps where the temperature depression condition is not fulfilled. This seems to suggest that the structure of the magnetic cloud is far more complex than anticipated.

The complex profile of both the sheath and the magnetic cloud is far from the ideal case of an unperturbed CME. It is not unreasonable to believe that this complexity could arise from an interaction between the partial halo CME launched on 13:25 UT 2012-01-18, and the halo CME erupting at 15:10 UT 2012-01-19. If so, it could be that the faster CME overtook the previous one, resulting in a front-sheath-front(-sheath) profile at Earth, which could very well explain the second temperature jump on 2012-01-22. Another possible explanation could be that the two CME fronts managed to merge, resulting in a single, enhanced IP shock front as seen at 06 UT on 2012-01-22, and an ongoing merging of the sheaths, resulting in a prolonged and messy sheath region. Since the 19 Jan CME event has been suggested to include the eruption of two flux ropes, and presumed that the previous eruption also included the eruption of a flux rope, it is apparent that the atypical MC profile could be due to the interaction of several flux ropes, further complicating the interpretation.

Nonetheless, it is an extremely difficult feat to derive the structure of a possible multiple ICME (MICME) event. In fact, since the temperature depression condition is still being fulfilled when the fast forward shock on 2012-01-24 arrives at Earth, we cannot exclude the possibility of a CIR interaction. In either case it is very likely that the error in SAT, as estimated by the model, is largely due to a multi-faceted plasmoid interaction in the heliosphere which could not be recreated by the model.

A look at the appended 3D simulation of the model, however, gives us a clue about the origin of the discrepancy between data and model. It seems like the two CMEs on the 18th and the 19th of January were expected to merge completely before reaching Earth, resulting in the single ICME profile as predicted by the model. If so, the data seem to suggest an unfinished merging of the ICMEs at the very least. The slight overestimation of the speed and temperature amplitude of the fast forward shock at 06 UT on 2012-01-22 seem to suggest a deficient match to the ambient solar wind. The discrepancy between the expected and the observed density amplitude of said IP shock is almost as to be expected, since the calculations of said parameters for halo CMEs are entwined with such large errors (see Section 2.1.3). The discrepancies between the expected magnetic field profile is a more serious matter, however, due to this parameter’s importance for potential geomagnetic effects.

5.2 2012 January 24

All plasma parameters were overestimated for this event, while the parameters related to the mag- netic field of the ICME was slightly underestimated, but not to such a degree as to yield a completely unexpected geomagnetic response. This in conjunction with the missing temperature depression and high variance seems to suggest that only the flank appears to have hit the ACE satellite. If so, the comparatively smaller discrepancy between the simulated and observed magnetic field profile seem to suggest a systematic underestimate of the magnetic field of ICMEs.

5.3 2015 March 17

The rise in the post-event speed compared to that prior to the IP shock, along with a change in the

opposite direction when looking at the densities of the entire time period, gives good support for