1 Introduction

Since their recognition as the primary causes of space weather storms (e.g. Wilson 1987; Gosling 1993; see Gopalswamy (2016) for a historical review), Coronal Mass Ejections (CMEs), and the Interplanetary Coronal Mass Ejection (ICME) disturbances that they produce in the solar wind, have been a focus of heliophysics research. Both the physical origins of CME events in the solar corona, and the resulting ICME plasma and field features in in-situ observations – here comprised of the leading shock and sheath compression together with the coronal material or “ejecta” – continue to be scientific targets of space mission investigations including the recently launched Parker Solar Probe (Fox et al.2016) and Solar Orbiter (Müller et al.2020). In addition, our understanding of the relationships of ICMEs to the CMEs observed by coronagraphs and extreme ultra-violet (EUV) imagers is particularly important because of its potential for forecasting the many local impacts of this form of solar activity (e.g. Russell 2000; Borovsky and Denton 2006; Burns et al.2007; Conner et al.2016; Reiff et al.2016). Progress has been significantly enhanced by the availability of multi-perspective imaging using combinations of Solar and Heliospheric Observatory (SOHO) and Solar Terrestrial Relations Observatory (STEREO) observations, starting in 2007. Those observations, together with STEREO/Heliospheric Imager (HI: Harrison et al.2018) white-light images that extended the field of view up to \(\approx 1.5\) AU, and spatially separated in-situ plasma and field measurements, reinforced the idea that the erupting structures seen in the corona become the drivers in their ICME counterparts. The results supported the results of earlier studies (e.g. starting with Burlaga et al.1981; Klein and Burlaga 1982; Marubashi 1986; Lepping, Burlaga, and Jones 1990) that envisioned expanding flux ropes, still partially attached to the Sun at both ends, plowing outward through the ambient solar-wind plasma and magnetic field – sometimes forming a leading shock as illustrated in Figure 1. This basic picture of an ICME and its coronal connection has been applied for decades to in-situ plasma and field measurement interpretations, heliospheric models, and geomagnetic-storm predictions that depend on knowing the arriving ICME speed and whether the structure includes southward (-Bz) interplanetary magnetic field (e.g. Gonzalez and Tsurutani 1987; Wilson 1987; Russell 2000). However, the multi-point perspectives and in-situ samplings tell us that the Sun-to-1 AU evolution is rarely the simple, self-similar flux-rope expansion picture suggested here.

Figure 1
figure 1

This “standard” picture of an ICME developed decades ago (e.g. Burlaga et al.1981; from Zurbuchen and Richardson 2006) includes the basic features that we assume today: the leading shock, a compressed ambient solar-wind “sheath” where the magnetic field may be perturbed by the shock and foreshock, and a shock “driver” consisting of the coronal “ejecta” including a flux-rope-like magnetic-field structure. Although this picture remains widely used, it leaves out some important details that continue to challenge our understanding and applications of this concept. (Image reproduced with permission from Zurbuchen and Richardson (2006), copyright by Springer.)

Full size image

Prediction of ICME properties at 1 AU from coronal observations of CMEs continues to be an elusive goal, as CMEs can deflect from a radial trajectory, can erode or merge with other structures, and can undergo non-uniform accelerations and decelerations in the course of their outward motion (see recent reviews by Manchester et al.2017; Kilpua, Koskinen, and Pulkkinen 2017; Cremades 2018). Simply connecting particular CMEs with the ICMEs that they produce is often difficult when the solar event signatures are not clearly observed or isolated from other activities. In addition, in-situ sampling is notoriously limited in its ability to provide a complete picture of an ICME’s large-scale structure. These complications, coupled with uncertainties regarding the CME initiation process(es) in the corona (see review by Green et al.2018, for example) have made it difficult to apply the early picture to both retrospective interpretations of well-observed events and potential geomagnetic-storm forecasting schemes. Thus, part of the strategy in designing the PSP and SO missions has been the prospect of revealing more insight into both the coronal roots of ICMEs and the roles played by their context and interactions as they evolve from coronal to interplanetary structures near the PSP perihelion of \(\approx 10~\mbox{R}_{\odot}\).

This overview provides a brief picture of ICME radial evolution between the corona and 1 AU as it is seen today. In Section 2 we revisit what has been learned based in large part on solar and heliospheric imaging, in Section 3 we consider the additional information gleaned from radio emissions, in Section 4 we summarize some key results from in-situ plasma, magnetic field, and energetic particle measurements, and in Section 5 we consider how far modeling efforts have been able to capture the full range of the observed ICME generation and propagation phenomena. Finally in Section 6 some thoughts on anticipated future contributions are offered. We end with an updated illustration of how the original concept in Figure 1 has evolved as a result of improved observations and interpretive modeling, and still-open questions to be addressed with the new perspectives and capabilities of the Parker Solar Probe (PSP) and Solar Orbiter (SO).

2 ICME Radial Evolution: As Revealed Through Images

Attempts to deduce the three-dimensional (3D) structure of coronal eruptions based on their appearance in two-dimensional images have been made for a long time (e.g. see Rouillard 2011; Webb and Howard 2012, and the references therein). Although there are sometimes narrow, “jet”-like transients observed in the corona (e.g. Vourlidas et al.2017; Sterling 2018), most CME “ejecta” near the Sun have been described as flux ropes (Mouschovias and Poland 1978; Chen et al.1997), ice-cream cone shaped bundles of coronal plasma and field (Fisher and Munro 1984), or spheromak-like flux toroids (e.g. Vandas et al.1997; Gopalswamy et al.2009a). Such models are still in vogue in various forms (e.g. Xie, Ofman, and Lawrence 2004; Xie et al.2006; Shiota and Kataoka 2016; Nieves-Chinchilla et al., 2018, 2019). Multi-perspective coronal imaging is now routinely used to determine the CME-ejecta orientations and propagation directions while they are still in the corona, assuming a croissant-like shape with its ends on the Sun. One of the main techniques is the graduated cylindrical shell (GCS) model (Thernisien, Vourlidas, and Howard 2009; Thernisien 2011), although there exist several alternatives (e.g. Isavnin 2016). Addition of a model for the initial ICME shock in the form of a spheroid surrounding the ejecta has made the picture of what is observed more complete (Olmedo et al.2013; Hess and Zhang 2014; Mäkelä et al.2015; Kwon, Zhang, and Vourlidas 2015; Xie et al.2017; Kwon and Vourlidas 2017). But how do these features evolve into their heliospheric counterparts? Are there features that are distinctly coronal phenomena? The coronal flux rope plus shock system is envisioned to undergo changes as it interacts with neighboring large-scale structures of coronal streamers and coronal holes (Gopalswamy et al.2009b; Wood et al.2012; Kay and Opher 2015; Liewer et al.2015), current sheets (Yurchyshyn, Abramenko, and Tripathi 2009; Isavnin, Vourlidas, and Kilpua 2014), other CMEs and ICMEs (Gopalswamy et al.2001a; Lugaz et al.2017a), and in the ambient solar wind with its spiral-shaped interacting stream structures (Gopalswamy et al.2001b; Savani et al.2010; Vršnak et al., 2013, 2014; Wang et al.2014). Work comparing the CME orientation in the corona to the ICME orientation at 1 AU often finds large deviations (Isavnin, Vourlidas, and Kilpua 2013; Wood et al.2017; Palmerio et al.2018), although it is not clear if these are due to CME rotation or limitations of the fitting and reconstruction methods (e.g. Al-Haddad et al.2013). Clear CME rotation is seen in coronal images (e.g. Kay and Opher 2015) but it is unclear how much more occurs in interplanetary space. In addition, different types of interactions dominate at different phases of the solar cycle (Gopalswamy, Tsurutani, and Yan 2015): Inter-CME and inter-ICME interactions dominate in the maximum phase because of the high rate of CMEs. At solar quiet times, the eruptions may be channeled toward the Equator when large polar coronal holes are present, while in the declining phase they may interact more with equatorial and mid-latitude coronal holes and their solar-wind streams. The outcomes of these interactions also depend on the intrinsic CME properties. A wide, fast eruption may evolve quite self-similarly, relatively unaffected by its surroundings, while a weak or even moderate event readily merges with the ambient solar-wind stream structure.

One of the primary motivations for coronal and heliospheric imaging studies is to determine whether the ICME shock strength, speed, and Bz (or north–south) magnetic-field component measured at 1 AU can be inferred from the CME structure and speed in the corona. While density and speed determinations from images and sequences of images, respectively, are relatively straightforward, the magnetic field determination from images has continued to present challenges. Although the fundamental structure of CMEs typically involves a magnetic-flux rope, they generally expand, distort and deflect as they interact with their surroundings en route to 1 AU. Their ICME counterparts, best represented by the in-situ signatures called magnetic clouds (MCs), are characterized by strong magnetic fields, smooth field rotations over large angles, low accompanying ion temperatures, and low plasma \(\beta \) (e.g. Klein and Burlaga 1982). Additionally, the handedness, or helicity, of the field in the active region from which a CME originates agrees with the apparent twist of the ICME ejecta fields for sometimes large fractions of studied samples (Cho et al.2013), as would be expected for well-matched CME/ICME pairs (e.g. Palmerio et al.2018). At 1 AU, only about 30% of ICMEs are MCs, but the statistics are generally considered to be influenced by the “observer” sampling of a generally present structure. In particular, only a fraction of the CMEs seen near the solar disk center, which are expected to have a direct impact at Earth, produce local MCs (Gopalswamy et al.2013a; Jian et al.2006; Jian, Russell, and Luhmann 2011; Vourlidas et al.2013; Li, Luhmann, and Lynch 2018). In the standard model for the initiation of CMEs, magnetic reconnection within closed, sheared coronal loops forms a flux rope and a post-eruption arcade (e.g. Forbes 2000; Fan and Gibson 2007). One of the consequences of the reconnection is that the heated plasma enters into the flux rope resulting in the presence of high charge states of minor ions inside MCs when observed at 1 AU (e.g. Lepri et al.2001; Lepri and Zurbuchen 2004; Reinard, Lynch, and Mulligan 2012). Analyzing a set of 54 CME–ICME pairs, Gopalswamy et al. (2013b) found that MCs and non-MCs were indistinguishable based on their near-Sun manifestations such as white-light CMEs and post-eruption arcades: the CMEs were fast and the flare arcades were well-defined (Yashiro et al.2013). Fe and O charge states at 1 AU were also indistinguishable between MCs and non-MCs, suggesting a similar eruption mechanism (Reinard 2008; Gruesbeck, Lepri, and Zurbuchen 2012; Gopalswamy et al.2013a). Furthermore, flux-rope fits to white-light CMEs by Xie, Gopalswamy, and St. Cyr (2013) revealed that MC and non-MC associated CMEs are on average deflected towards and away from the Sun–Earth line, respectively. The different deflections of MC and non-MC CMEs was further confirmed by the different coronal-hole influence parameters for the two groups of CMEs (Mäkelä et al.2013). Thus, the deflection away from the Sun–Earth line of the non-MC CMEs is consistent with the view that the observing spacecraft often pass through their flanks, missing the central flux-rope structures and resulting in their non-MC appearance (see, e.g., Marubashi 2000; Owens et al.2005; Gopalswamy 2006a; Jian et al.2006; Jian, Russell, and Luhmann 2011; Kim et al.2013). Marubashi et al. (2015) showed that almost all ICMEs can be fit with a flux-rope model if a locally toroidal (e.g. spheromak) model is also considered in addition to the traditional cylindrical picture. This form may be especially suitable for ejecta crossings far from their centers.

The majority of ICME studies involving heliospheric images has focused on understanding their interplanetary propagation toward determining their direction or forecasting their hit/miss and/or arrival times (Wood and Howard 2009, Liu et al.2010; Lugaz et al.2010; Rouillard 2011; Wood et al.2011; Möstl et al., 2014, 2017). A few of these have also tried to take into account their radial expansion (Savani et al., 2009, 2012; Lynch et al.2010; Nieves-Chinchilla et al.2012; Lugaz et al.2012), interplanetary evolution (Poomvises, Zhang, and Olmedo 2010), and deformation (Savani et al., 2011a, 2011b). Savani et al. (2015) projected a reconstructed local structure back onto the solar surface to determine the central axis of the initial CME, its source region, and related coronal magnetic structure, and they compared that information to the in-situ ICME magnetic structure detected upstream of Earth. Another method for connecting the magnetic structure of the solar CME to the ICME (Gopalswamy et al.2018a) is the “flux rope from eruption data” (FRED) technique. This approach infers the total reconnected flux in the eruption region (Qiu and Yurchyshyn 2005; Qiu et al.2007; Hu et al.2014; Gopalswamy et al.2017a), and then assumes self-similar expansion to 1 AU (Gopalswamy et al.2018b) to estimate arrival time and predicted Bz-component (Scolini et al.2019; Singh et al.2019). In a particularly comprehensive CME image/ICME analysis, Wood et al. (2017) focused on the ability to infer the arriving MC’s Bz-component sign – essential for geomagnetic-storm predictions. They investigated whether routinely fitting classical croissant-shaped CME flux-rope pictures to multi-perspective coronal and heliospheric images (as illustrated by the magenta field lines in Figure 2), and determining the photospheric fields at their roots, could be used to distinguish the resulting ICMEs that had geo-effectively important southward Bz at 1 AU. Similar techniques have been tested by Kay, Opher, and Evans (2015), Möstl et al. (2018), and Palmerio et al. (2018). Nieves-Chinchilla et al. (2018) and Al-Haddad et al. (2019) concluded, as Wood et al. (2017) had, that this kind of projection procedure is not widely applicable to connecting the structures observed at the Sun and the magnetic fields of the ICMEs at 1 AU (also see Kilpua et al.2019).

Figure 2
figure 2

Example illustrating a spherical coronal shock (yellow grid) and coronal flux rope (magenta field lines) fit to multi-perspective coronal images to interpret, visualize, and empirically model the initial phase of what becomes the ICME in the heliosphere (NASA image).

Full size image

The most energetic solar-particle events, such as those that produce ground-level enhancement (GLE) events on Earth (Mewaldt et al.2012), are inferred to have their sources close to the Sun. Thus, there has been significant attention to the question of the onset and evolution of the interplanetary shocks associated with CMEs and ICMEs. This includes the challenge of separating CME and shock structures so that their relationship and consequences can be studied. Their relative 3D evolution affects the interpretation of solar energetic-particle distributions and intensities throughout the inner heliosphere, as well as the coronal dimmings and EUV waves observed in the low corona (e.g. Ma et al.2011) in conjunction with the early stages of CME-to-ICME transition. Kwon, Zhang, and Olmedo (2014) used SOHO and STEREO’s \(\approx 360\)° coverage of the corona to investigate the 3D structure of CME-driven shocks and how they are related both to the EUV waves and the CME leading edges in white-light images. A simple geometric representation of the shock as a sphere was found to provide a good description of the coronal disturbance surrounding the erupted material in all three coronagraph viewpoints. This work supported the earlier suggestions that the halo appearance of some CMEs can at times be due to the shocks rather than the CMEs themselves (Shen et al.2014), and that the EUV wave is the low coronal reaction to the initial expanding CME disturbance (see also Attrill et al.2007; Downs et al.2011; Long et al.2017). In limb events, EUV waves are often interpreted as shock footprints (Veronig et al.2010; Kozarev et al.2011; Ma et al.2011; Patsourakos and Vourlidas 2012; Gopalswamy et al.2012a). More recently, Kwon and Vourlidas (2017, 2018) refined the methods of fitting these coronal shocks and extracting information about their strength (e.g. their density compression ratios), important to understanding their potential for accelerating solar energetic particles (SEPs) in the corona early in the overall event. The knowledge, from images, of the shock location in the early stages of CME liftoff allows researchers to study the relationship between SEPs arriving at multiple heliospheric locations and their coronal source as done by Rouillard et al. (2011) and Lario et al. (2014, 2016, 2017). Those authors found mixed results in using observer field-line backward mapping to the coronal-shock signatures, but the insights gained regarding both the interpretations of the observations close to the Sun and their connections to what is observed at 1 AU are important.

The detectability of high-energy particles at a given observer location (e.g. associated with a GLE) may depend on its magnetic connectivity to the nose of the shock (where it is expected to be strongest) determined from images of the early stages of CMEs in white light or EUV (Gopalswamy et al., 2014a, 2016). When STEREO observations with the extended coronagraph field of view close to the solar surface (STEREO/COR1) are available together with EUVI images, it is possible to accurately determine the shock-formation height (Gopalswamy et al., 2009c, 2013d): an important parameter to determine the particle acceleration efficiency of CME-driven shocks (Gopalswamy et al.2017b). The shock formation depends on the relative importance of the Alfvénic (or magnetosonic) speed profile in the corona and the CME-speed profile (Gopalswamy et al.2001c; Mann et al.2003). As mentioned earlier, in GLE events, the inferred shock-formation height is typically around \(1.5~\mbox{R}_{\odot}\) (Gopalswamy et al., 2013e, 2018c; Thakur et al.2014) and is consistent with the velocity dispersion analysis of GLE SEPs (Reames 2009). On the other hand, the shock-formation height is much larger (about 5 – \(10~\mbox{R}_{\odot}\)) in the case of accelerating CMEs associated with filament eruptions outside of active regions (e.g. Kahler 2001; Gopalswamy et al., 2015c, 2016, 2017b). The particle-acceleration efficiency is also determined by the speed of the CME at the shock-formation height. If the shock-formation height is low and the CME speed is also low, then one gets only small SEP events (Gopalswamy et al.2017b). If the observer is connected to the weaker shock flank, the resulting particle spectrum becomes softer (Gopalswamy et al.2018c).

The coronal shock is expected to form at a stage when the CME/ICME evolution is dominated by the heating and magnetic-pressure-induced expansion associated with the combined flare and CME eruption, and it weakens as the local effects subside. As the expanding ejecta move outward, the less-symmetrical driven shock takes over. The evolution of the shock standoff distance is an indication that the driving CME slows down with distance from the Sun (Gopalswamy and Yashiro 2011). Its efficiency as an SEP source is not determined, although SEPs may also be accelerated in the flare and eruption processes. But the “driven” shock can be maintained by the outwardly moving and expanding ejecta from the corona to large radial distances and is a source for large “gradual” SEP events that can last several days up to, and beyond, the ICME shock arrival time at 1 AU (Cohen 2006; Mäkelä et al.2011; Reames 2017). As already mentioned, GLE timing suggests that the shock must already have formed at less than \(10~\mbox{R}_{\odot}\) heliocentric distance, which seems to be the case for large SEP events in general (Ma et al.2011; Kozarev et al.2015; Gopalswamy et al.2017b). Note that some driven shocks identified near the Sun using Type-II burst observations may not arrive at 1 AU for various reasons (Gopalswamy 2006b; Gopalswamy et al.2012b). The driving CMEs are of lower energy, so the shocks may dissipate before arriving at Earth, or closely spaced and timed CME shocks may merge, resulting in a single shock at Earth (Schmidt and Cargill 2004; Lugaz, Manchester, and Gombosi 2005), or deflection of the shock driver or the shock itself, away from their original Sun–Earth line trajectory by nearby coronal and solar-wind structures may occur. However, the probability of observing the shock at 1 AU increases rapidly when the imaged CME speed exceeds \(1000~\mbox{km}\,\mbox{s}^{-1}\) and when the Type-II bursts are observed down to frequencies below 1 MHz. Radio observations of particle acceleration regions around CMEs are still increasing in their capability (e.g. Zucca et al.2018; Morosan et al.2019), and are expected to reveal further details of where and when CME and ICME related sources become important.

The present Cycle 24 solar activity went through its (comparatively weak) maximum in \(\approx 2011\) – 2014 (e.g. Gopalswamy et al.2015b) when the multipoint, multi-perspective observational resources provided by \(\mbox{L}_{1}\) and STEREO spacecraft were all available. During this period, it became better appreciated that the larger, faster CMEs can occur in quick succession and can also be accompanied by lesser eruptions (Möstl et al.2012; Lugaz et al.2012; Gopalswamy et al.2013c; Liu et al.2014a, 2014b; Temmer et al.2014), confirming findings made with LASCO during Solar Cycle 23 (Gopalswamy et al.2001a). As a result, the possibility of CMEs interacting in the inner heliosphere significantly increases. CME–CME interactions are important both because they affect CME/ICME propagation and evolution, and, from a space-weather point of view, because the resulting geomagnetic and SEP effects can be greatly altered (see review by Lugaz et al.2017a). SOHO and STEREO studies of interacting CMEs, with the addition in 2010 of the Solar Dynamic Observatory (SDO)’s EUV imaging capabilities, have taken advantage of the combined solar observations and separated in-situ measurements to understand their detailed characteristics. For example, Figure 3 (from Maricic et al.2014) shows an example where HI observations recorded the merging of three CMEs near the Sun that eventually impacted the Earth and caused a single complex response. STEREO observations of Earth-directed CMEs revealed small CMEs preceding SOHO/LASCO halo CMEs that helped explain the altered ICME travel times (Gopalswamy et al.2013c). Webb et al. (2013) described the complicated structure of the inner heliosphere during a very active period in early August 2010, an interval that was also the subject of dedicated studies and publications (Temmer et al.2012; Harrison et al.2012; Möstl et al.2012). Some authors reconstructed 3D heliospheric densities from the images and compared them with the timing and magnitude of in-situ density structures at five spacecraft locations spread over 150° in ecliptic longitude and 0.4 to 1 AU in radial distance, together with modeled local flux-rope structures (Webb et al.2013). This work highlighted the difficulties in using kinematics to describe the morphological evolution of ICMEs during periods of widespread activity. Many of the intense geomagnetic storms from Solar Cycle 24 resulted from the succession or interaction of CMEs (Gopalswamy et al.2015b; Lugaz et al.2016; Shen et al.2018). CMEs and ICMEs always interact with ambient flows of different origins as they propagate from the corona into the interplanetary medium. White-light image-based “J-maps” suggest that interaction with large-scale structures close to the Sun, including coronal streamers, coronal holes, solar-wind stream boundaries, and other CMEs, can severely alter the shapes and trajectories of the coronal ejecta and affect the ICMEs they evolve into. Their ongoing interactions during ICME propagation impact the plasma and field parameters associated with the arriving plasma and field disturbance(s), and the associated SEPs.

Figure 3
figure 3

STEREO/HI images of three CMEs merging near the Sun and propagating outward as a complex structure. The dashed line is the Ecliptic. The o- and x-points mark leading-edge features of two of the original events. (Image reproduced with permission from Maricic et al. (2014), copyright by Springer.)

Full size image

The physical details of the encounters between CMEs/ICMEs is complicated and determines whether their interaction is constructive, destructive, or neutral. A recent review of the subject by Lugaz et al. (2017a) discusses a number of examples and studies. Early work based on heliospheric images (Shen et al.2012; Lugaz et al.2012) found evidence of both deflection of one CME/ICME by another and of super-elastic collisions, in which momentum redistribution occurs in ways that enhance the event. Several investigations examined the kinematics of ICME interactions to determine the manner in which they merge. Maricic et al. (2014) described a chain of events on 14 – 15 February 2011 and found evidence for a gradual momentum transfer from the faster to the slower CME ahead. They inferred that momentum transfer may result from alteration of the following events’ shock propagation as they travel through the preceding events. This was interpreted as causing additional drag on the faster ICMEs, resulting in deviations from expected travel times (see also Gopalswamy et al.2013c). Temmer et al. (2014) presented a detailed analysis of the interaction of two of these CMEs. It was found that the interaction process strongly depends on the geometry, with differences in the outcome for interacting ICME flanks versus apexes, and the most centrally located interaction showing the strongest changes in kinematics. This topic was also addressed by Mishra and Srivastava (2014), who described evidence in the in-situ observations near 1 AU of acceleration, compression, and heating of the leading ejecta. Mishra, Wang, and Srivastava (2016) studied a case involving two interacting CMEs that occurred on 25 October 2013. They considered the propagation and expansion speeds, impact geometry, angular sizes, and masses of the interacting CMEs using 3D-reconstruction techniques applied to STEREO/SECCHI-COR and -HI observations, and they found that the higher expansion speed of the following CME compared to the preceding CME may increase the probability of interaction. In general, interactions with their surroundings can seriously affect both the initial state and evolution of ICMEs in the interplanetary medium. In particular, when there are small-to-moderate CMEs ahead of a large CME that is potentially geo-effective, the arrival time of the larger ICME can be significantly altered relative to the expected time based on the coronal observations. One attempt has been made to develop an analytical model to determine where CMEs interact near the Sun, and the consequences for their arrival time at Earth (Niembro et al.2015). The approach was tested on several real events, with results that seemed to depend on how well they could be described and tracked. However, just the knowledge that there have been significant interactions can be important in both event interpretations and space-weather forecasting.

An exceptionally fast and large CME seen as a halo event on STEREO-A when the spacecraft was about 120° ahead of Earth on 23 July 2012 inspired excitement in part due to its occurring during the relatively weak Solar Cycle 24. Its near-Sun speed, at upwards of \(\approx 2500~\mbox{km}\,\mbox{s}^{-1}\), was at the high end of observed CME speeds. The event also gave rise to an exceptionally intense SEP event detected by STEREO-A (Russell et al.2013), where the leading shock of the ICME was likely eroded by the significant local pressure contribution from the large density of SEPs. The question of how such extreme space-weather storms are born and evolve, and how severe they can be when they reach Earth, was examined by Liu et al. (2014a), where the authors investigated this period using multi-point remote-sensing and in-situ observations. At least three effects of multiple-event interactions were found to influence what was observed: i) deflections in the propagation direction of the fastest CME/ICME leading to a head-on impact with STEREO-A, ii) extreme compressions of the ejecta magnetic fields to over 100 nT, and iii) minimal deceleration during the Sun-to-STEREO-A transit due to a prior event having left a rarefied ambient solar wind in its wake (Liu et al.2014a; Temmer and Nitta 2015). Together, these caused an unprecedented set of interplanetary conditions, including an extended period of strong southward magnetic field (-Bz) which, had it arrived at Earth, would have produced a record geomagnetic storm. Gopalswamy et al. (2016) compared the 10 – 100 MeV proton spectrum of this event with other large SEP events and found that the 23 July 2012 event was similar to those with GLEs, meaning that GeV particles were likely accelerated. This result also implies that the CME attained very high speeds close to the Sun. These results provided new insights on how an extreme space-weather event can arise from a combination of conditions and events both at the Sun and during propagation to 1 AU.

3 ICME Radial Evolution: As Revealed by Radio Emissions

Solar radio emissions provide a unique remote-sensing diagnostic of the CME process, and its related shock generation and propagation. One of the radio burst types closely related to the occurrence of CMEs is the Type-II burst (see Nelson and Melrose 1985, for a review). Type-II bursts have been observed in the frequency range from 100s of MHz to tens of kHz (Cane and Stone 1984; Cane, Sheeley, and Howard 1987; Gopalswamy et al.2012a; Cho et al.2013). From a plasma-frequency perspective, these frequencies span the spatial domain starting from the inner corona to the vicinity of observing spacecraft at \(\mbox{L}_{1}\), and occasionally even further. This provides the opportunity to track shocks throughout the inner heliosphere starting from about \(1.1~\mbox{R}_{\odot}\) to \(>215~\mbox{R}_{\odot}\) (e.g. Gopalswamy 2011; Liu et al.2013; Cremades et al.2015). When the Radio and Plasma Wave Experiment (WAVES: Bougeret et al.1995) onboard the Wind spacecraft became available, observations in the decameter–hectometric (DH) wavelength domain led to important discoveries regarding CME interactions (Gopalswamy et al.2001a) and the establishment of the relationship between CME kinetic energy and the wavelength range over which the radio emission takes place (Gopalswamy et al.2005).

Figure 4a shows a schematic dynamic spectrum with the slanted lines indicating Type-II bursts in various wavelength ranges. Some Type-II bursts start and end in the metric domain; some start in the metric domain and continue to be present in the kilometric domain. There are also intermediate cases. Finally, some bursts start in the kilometric domain. The Type-II wavelength range provides important information about the CME kinematic evolution in the corona and interplanetary (IP) medium. Purely metric Type-II bursts are associated with CMEs that have an average speed of \(\approx 600~\mbox{km}\,\mbox{s}^{-1}\). CMEs producing Type-II bursts in the DH domain have an average speed of \(\approx 1100~\mbox{km}\,\mbox{s}^{-1}\). Type-II bursts with emission components at all wavelengths (metric, DH, and kilometric) are produced by the fastest CMEs (\(\approx 1500~\mbox{km}\,\mbox{s}^{-1}\)) (Gopalswamy 2011). CMEs producing purely kilometric Type-II bursts have the lowest average speed (\(\approx 550~\mbox{km}\,\mbox{s}^{-1}\)), only slightly smaller than that of CMEs associated with metric Type-II bursts. All CMEs, except those associated with kilometric Type-II bursts, have an average deceleration in the coronagraph field of view. CMEs associated with the kilometric Type-II bursts have average positive acceleration and attain super-Alfvénic speeds at several tens of solar radii from the Sun where they form shocks, and they produce the radio emission at long wavelengths. The slope of the lines [\(\mbox{d}f/\mbox{d}t\)] in Figure 4a is related to the shock speed and the density scale height. The burst drift-rate spectrum (Figure 4b) provides a picture of the shock evolution from the corona to the interplanetary medium. In the inner corona, the drift-rate spectrum is flat, indicating an accelerating source, while in the interplanetary medium, the drift-rate spectrum is steep, indicating deceleration (see Gopalswamy 2011, for details).

Figure 4
figure 4

(\(\mathbf{a}\)) Type-II bursts occur at different wavelength ranges, as illustrated here: 1. purely metric, 2. metric to DH, 3. DH, 4. DH – kilometric, 5. metric to kilometric, and 6. purely kilometric. (\(\mathbf{b}\)) The drift rate [\(\mbox{d}f/\mbox{d}t\)] dependence on the emission frequency [\(f\)] in the m, DH, and km domains using data from various sources (from Gopalswamy 2011) indicates whether the source is accelerating or decelerating.

Full size image

In a recent work, Gopalswamy et al. (2018d) tracked the shock speed from the corona to IP medium, using a well-observed Type-II burst during the 21 June 2015 CME, which produced the second largest geomagnetic storm in Cycle 24 and a large SEP event (Liu et al.2015). The Type-II emission was observed from metric to kilometric wavelengths. The 3D speed of the CME was determined by fitting a flux rope to the SOHO/LASCO observations. The leading-edge speed of the CME was tracked in the coronal images and using the Type-II burst that was also observed near the Wind spacecraft when the shock arrived. The drift rate of the Type-II burst was determined at several heliocentric distances and the shock speed was derived. The combined coronagraph, Type-II burst, and in-situ observations provided the complete record of the evolution of the shock speed from the Sun to Earth shown in Figure 5. The combined data set captured the complete evolution: the initial rapid increase, slow increase in the outer corona, rapid decline within \(\approx 100~\mbox{R}_{\odot}\), and finally a slower decline until the shock was detected in situ by the Wind spacecraft. Using the CME images, it was also possible to deduce that the metric Type-II emission originated from the flanks of the shock approximately 60° from the nose, while the IP Type-II burst originated from the nose region of the shock. This example demonstrates how the radio signatures of ICME shocks can provide substantial and unique additions to what is obtained from the imaging observations regarding the evolution scenario interior to 1 AU.

Figure 5
figure 5

Sun-to-Earth evolution of a shock associated with the 21 June 2015 CME. SDO and SOHO observations provided the open circles, while the squares are from the Wind/WAVES radio dynamic spectrum. The in-situ shock speed at \(\mbox{L}_{1}\) from Wind (\(776~\mbox{km}\,\mbox{s}^{-1}\)) is shown by the solid circle. (Image reproduced with permission from Gopalswamy et al. (2018d), copyright by Elsevier.)

Full size image

In a study involving 222 IP shocks detected by Wind and/or Advanced Composition Explorer (ACE), Gopalswamy et al. (2010) found that about \(\approx 34\%\) lacked Type-II radio bursts (radio-quiet or RQ shocks). The CMEs associated with the RQ shocks were generally slow (average speed \(\approx 535~\mbox{km}\,\mbox{s}^{-1}\)) compared to those associated with radio-loud (RL) shocks (average speed \(\approx 1237~\mbox{km}\,\mbox{s}^{-1}\)). The average Sun-to-Earth transit speeds of RQ and RL shocks were \(629~\mbox{km}\,\mbox{s}^{-1}\) and \(851~\mbox{km}\,\mbox{s}^{-1}\), respectively. This is consistent with the lower CME kinetic energy associated with RQ shocks. CMEs associated with RQ shocks were generally accelerating within the coronagraph field of view (average acceleration \(\approx +6.8~\mbox{m}\,\mbox{s}^{-2}\)), while those associated with RL shocks were decelerating (average acceleration \(\approx -3.5~\mbox{m}\,\mbox{s}^{-2}\)). This means that many of the RQ shocks formed at large distances from the Sun, typically above \(10~\mbox{R}_{\odot}\), consistent with the absence of metric and DH Type-II radio bursts. A Type-II burst starting at a frequency of 300 kHz (1 km wavelength) indicates that the shock forms at a distance of \(\approx 20\) solar radii as inferred from a simple density model (e.g. Leblanc, Dulk, and Bougeret 1998). Longer-wavelength bursts imply shock formation at even larger distances from the Sun. The Alfvénic Mach numbers of RQ shocks at 1 AU average 2.6 compared to 3.4 for RL shocks, suggesting that RQ shocks were mostly subcritical, so they were not efficient in accelerating electrons (hence radio quiet). About 18% of the in-situ shocks studied do not have discernible ejecta behind them. These shocks are probably due to CMEs moving at large angles from the Sun–Earth line as mentioned in the earlier discussion of observer sampling geometry, although some could also be associated with solar-wind stream interaction regions. As with all of the observing techniques discussed here, the overall event geometry relative to the observer determines how much information can be extracted.

Radio enhancement signatures associated with CME interactions provide another opportunity, in addition to the images, to diagnose these occurrences with remote sensing. These have been interpreted as the signature of the acceleration of additional electrons (Gopalswamy et al.2001a). Mäkelä et al. (2016) were able to identify the location of this acceleration using the direction-finding technique from STEREO and WAVES observations. The interacting CMEs on 2 May 2013 observed by SOHO and STEREO had a radio-enhancement source that was located at the interface between the two interacting CMEs. It has been argued that the same may be happening for protons: major SEP events may result from these CME interactions occurring close to the Sun where the CME-driven shocks are the strongest (e.g. Gopalswamy et al., 2002, 2004; Li et al.2012). Recently, Ding et al. (2019) studied 64 radio-enhancement events and the associated SEP events from Solar Cycle 24. They confirmed that the radio-enhancement signature is the key difference between SEP-rich and SEP-poor eruptions, as suggested by Gopalswamy et al. (2002). This is another piece of information that complements the imaging information toward inferring what is happening as the ICME evolves close to the Sun.

4 ICME Radial Evolution: As Revealed by In-Situ Measurements

The in-situ features of ICMEs have been analyzed by many authors whose goals have ranged from understanding the structure of these interplanetary disturbances, to making connections to their solar counterparts, to considering their role as the external drivers of geomagnetic storms. Figure 6 contains examples of the 1 AU in-situ signatures of a few large ICMEs observed in the STEREO era, including the plasma, magnetic field, and SEPs (also see Jian et al., 2006, 2018d). Although the 1 AU measurements, like the coronal images, indicate the presence of erupted coronal material (including magnetic-flux ropes) preceded by a shock (for fast-moving cases) and sheath, their relationship to what is observed at the Sun is not totally clear (e.g. Dasso et al.2007; Manchester et al.2017 and the references therein). Focusing on the 1 AU ICMEs associated with observed CMEs, Möstl et al. (2014) examined 22 events seen both at the Sun and in heliospheric images, and at various 1 AU locations, to give a more comprehensive empirical picture of how the kinematics, directions, and global shapes of ICMEs change during propagation from the Sun. Li, Luhmann, and Lynch (2018) similarly analyzed the relationships between a large selection of in-situ events that had magnetic-cloud drivers, identifying the solar signatures that seemed to be most closely related – including flares, filament eruptions, and “stealth” types of sources, where no solar counterpart could be identified in the latter (e.g. Howard and Simnett 2008). Nieves-Chinchilla et al. (2013) used joint in-situ/imaging observations to study the 3D evolution of a particular stealth CME in detail. An important finding in several of these studies was that the errors of hours, and sometimes days, in expected arrival times based on the coronal CME time and speed was difficult to account for, with possible contributors including the history of interactions with ambient structures and the assumed shape and flux content of the propagating driver. On the other hand, He et al. (2018) tracked the 8 October 2016 stealth CME with SDO, SOHO, STEREO, and in-situ observations, and, using additional modeling, they were able to infer that the stealth CME was bracketed between slow solar wind ahead and a fast stream behind. In general, an essential part of interpreting what is observed is understanding the surrounding plasma and field context of the event, and its physical interactions with those surroundings as it travels (also see Farrugia et al.2011). Processes referred to as drag and/or erosion (e.g. Ruffenach et al., 2012, 2015) are often not explicitly considered in event analyses, and are still poorly understood. These studies represent only a small fraction of recent work on in-situ studies of ICMEs. Many others are described in the comprehensive review by Kilpua, Koskinen, and Pulkkinen (2017) and the references therein.

Figure 6
figure 6

Examples of ICMEs observed with in-situ instrumentation on STEREO. The panels show (from top to bottom) proton temperature, density (both showing the ICME sheath compression and heating of solar wind at the leading shock marked by dashed lines), plasma velocity, magnetic-field magnitude, and north–south (Bz) component showing the enhancement and rotations associated with the magnetic-cloud driver, and the SEP protons that are the first arriving in-situ signature.

Full size image

The solar-cycle dependence of ICMEs has been well-documented. Although their occurrence rate trends are not always a reflection of the related sunspot-cycle sizes, during the relatively weak Solar Cycle 24 ICMEs have occurred less often and have been generally weaker and slower than in the previous cycle (Gopalswamy et al.2015b; Chi, Shen, and Wang 2016; Jian et al.2018d). Fewer CMEs were also seen in the coronal observations by Hess and Colaninno (2017), especially the fast and wide ones that usually result in easily identifiable ICMEs. Gopalswamy, Tsurutani, and Yan (2015) suggested this decrease in ICME numbers was not solely due to propagation differences and/or ICME size and identification issues. Although the mean speeds of the CMEs close the Sun are similar for Cycles 23 and 24 (Gopalswamy et al.2014b), the measured ICME speeds at 1 AU in Cycle 24 are slower. Reduced total pressures near the Sun, where coronal densities and fields have diminished relative to earlier cycles, may have allowed the Cycle 24 CMEs to expand more quickly close to the Sun (Gopalswamy et al.2015a), while the ICME widths at 1 AU decreased (Jian et al.2018d). One also cannot rule out a possible in-situ sampling bias in the impact parameter that is different for the two cycles, perhaps due to greater average deflection of the eruptions from the ecliptic plane (e.g. Kay, Opher, and Evans 2015).

The MCs mentioned earlier represent a significant subset (\(\approx 30\%\)) of observed ICMEs at 1 AU, especially during less active periods surrounding solar minimum (e.g. Richardson and Cane 2004). As a major factor affecting the related geomagnetic activity, the north–south or Bz-component of these ejecta exhibit bipolar (Bz changes sign) and unipolar (Bz does not change sign) configurations when the axis of the flux rope is oriented at low \(vs\). high inclinations with respect to the ecliptic plane, respectively (e.g. Mulligan, Russell, and Luhmann 1998; Gopalswamy 2008). Li et al. (2014) and Li, Luhmann, and Lynch (2018) have made long-term analyses of the solar-cycle trends in the polarity of bipolar MCs, defined as the time-ordering of the rotation, which varies with the Hale cycle as illustrated in Figure 7. While bipolar MCs show a clear inter-cycle variation, the unipolar MCs do not show such a variation, at least in Solar Cycle 23. However, Li, Luhmann, and Lynch (2018) and Nieves-Chinchilla et al. (2019) found there were more unipolar MCs at solar maxima although the preference of north or south orientation does not have any clear solar-cycle dependence. From Solar Maximum 23 to Solar Maximum 24, the NS (north to south) MCs dominated, and presently the SN (south to north) MCs dominate (Li, Luhmann, and Lynch 2018; Nieves-Chinchilla et al.2019). Li, Luhmann, and Lynch (2018) found the solar-cycle dependence of the bipolar MC average polarity is attributable to MCs that originate from quiescent filaments in decayed active regions, while those associated with flaring active-regions have mixed Bz polarity without solar-cycle dependence.

Figure 7
figure 7

Results from analyzing the magnetic “polarity” (north–south or Bz magnetic field) of several decades of in-situ observations of ICME magnetic-cloud drivers. The clear solar-cycle trends in the occurrence rates of magnetic clouds with leading northward (NS) and leading southward (SN) magnetic-field rotations are seen here, including the effect of the solar polar-field polarity on which dominates. Adapted from Li, Luhmann, and Lynch (2018).

Full size image

Significant progress on the subject of ICME transport and evolution was made possible over the maximum of Solar Cycle 24 by the solar-activity level and the availability of spacecraft at various locations in both radius and helio-longitude: MErcury Surface, Space ENvironment, GEochemistry, and Ranging (MESSENGER) at \(\approx 0.3~\mbox{AU}\), Venus Express at \(\approx 0.7~\mbox{AU}\), and Wind, ACE, SOHO, SDO, and STEREO with its dual, separated perspectives at 1 AU. In particular, the multipoint in-situ observations from 0.3 to 1 AU and multi-perspective images from 1 AU provided an essential link for interpreting solar-system-wide evolution and consequences of both solar-wind structure and ICMEs. STEREO often provided the closest 1 AU location for in-situ data comparison for planetary missions because of its longitudinal separation from the Sun–Earth line. For the first time since the mid-1980s with the Helios mission, MESSENGER provided in-situ measurements (although with limited plasma data) at heliocentric distances of less than 0.5 AU, often within the field-of-view of STEREO’s Heliospheric Imagers. In addition, the relatively fast orbit of Mercury (88-day orbital period) resulted in many potential conjunction events with solar-wind monitors near 1 AU (Wind/ACE and STEREO).

Planetary scientists have made a wide variety of uses of the distributed heliospheric measurements to interpret space-weather conditions at Mercury, Venus, Mars, and beyond including studies of the radial evolution of these transient structures. Winslow et al. (2015) assembled a list of ICMEs that impacted MESSENGER while it orbited Mercury (2011 – 2015). They found that the average ICME expansion between 0.3 and 1 AU was very similar during Solar Cycle 24 to that inferred from Helios measurements during Solar Cycle 21. They then used STEREO and \(\mbox{L}_{1}\) data together with MESSENGER observations to evaluate the statistical differences in ICME strengths between the orbital distances of Mercury and Earth. Their analysis also showed the overall weakening of most leading shocks, consistent with an average deceleration of the ejecta drivers in that heliocentric-distance range. In another study, Winslow et al. (2016) used conjunctions of STEREO-A and MESSENGER to investigate the development of complexity during the propagation of several ICMEs between the two locations, although Good et al. (2015) had found that some events simply expand self-similarly over this radial range. The ambient conditions, e.g. whether significant solar-wind stream structure and/or the heliospheric current sheets are present, appear to influence the outcome. Good and Forsyth (2016) assembled a catalogue of MESSENGER ICMEs covering its seven-year transit to Mercury with Venus Express (VEX), STEREO, and ACE event data, enabling the determination of ICME widths at different radial distances and their consistency with self-similar expansion (Good et al.2019). They identified 23 ICMEs observed by pairs of spacecraft in close radial alignment, providing a valuable resource to those seeking to analyze the differences in space weather at the terrestrial planets and the heliospheric contexts of events occurring during the MESSENGER and VEX missions (2006 – 2013). A study that incorporated both Mars Science Laboratory (MSL)/Radiation Assessment Detector (RAD) and Mars Express data (Möstl et al.2015) used STEREO observations to model the direction and expansion of a Sun-to-Mars event, concluding that non-radial propagation can be significant. Witasse et al. (2017) presented a case study that may have included ICME detection as far out as Saturn, by the Cassini spacecraft, and at New Horizonsen route to Pluto. However, the level of detail derivable from the collection of observers in this case mainly makes the point that those ICMEs that propagate well beyond 1 AU can still be identified with a particular solar event. Jian et al. (2006, 2008a, 2008b) and Jian, Russell, and Luhmann (2011) conducted a number of analyses focusing on determining the changing properties of ICMEs at Venus, Earth, and Jupiter orbital distances using Pioneer Venus Orbiter (PVO), Wind/ACE, and Ulysses data, respectively, with sufficient events in each location to determine solar-cycle variations. Their results show that, as the ICMEs propagate away from the Sun, they increasingly interact with other ICMEs or solar-wind stream interaction regions, with the fraction as high as 37% at 5.3 AU (Jian et al.2008b). From 1 to 5.3 AU, the occurrence of ICME shocks decreases slightly and the expansion rate of the ICME drivers is less than it is within 1 AU (Jian et al.2008b).

Most recently, Janvier et al. (2019) derived average properties and magnetic-field temporal profiles of ICMEs for Mercury, Venus, and Earth heliocentric distances from the collected observations, finding radial trends that generally agree with the conclusions of the multipoint case studies, but also adding insight regarding evolution of observed asymmetry in the ICME temporal profile. As asymmetries are affected by impact parameter sampling of the passing structure, which can introduce apparent asymmetry, interpretation is challenging. Asymmetries can also result from propagation effects. Dasso et al. (2006) pointed out that magnetic reconnection at the front and/or rear edges of ICME flux ropes would reduce the magnetic flux of a flux rope and alter its observed polarity pattern. Ruffenach et al. (2015) selected 50 MCs and found that nearly 30% of them showed potential reconnection signatures at their boundaries, with average erosion of about 40% of the total azimuthal magnetic flux. It is clear that the ICME propagation and evolution processes, as well as sampling considerations, make it difficult to interpret in-situ observations in a straightforward way.

A topic of special interest in space-weather forecasting concerns the ability of the ICMEs to drive shocks and the heliocentric distance of shock formation. ICME ejecta both expand and propagate, leading to shock formation in cases where they do so with supermagnetosonic speeds relative to the background plasma (Siscoe and Odstrcil 2008). In the corona, strong events can accelerate particles within mere minutes after their initiation (e.g. Gopalswamy et al.2012c). On the other hand, there are, on average, more CMEs with shocks at 1 AU than at 0.7 AU, from 48% to 65% although this result comes from analyses of long-term observations obtained during different solar cycles (Jian et al.2008a, 2008c). As mentioned earlier, radio-quiet shocks (Gopalswamy et al.2010; Janvier, Démoulin, and Dasso 2014), and shocks with purely kilometric Type-II bursts (Gopalswamy et al.2005) are known to form at large distances from the Sun and hence may provide an explanation for the higher abundance of shocks at 1 AU. Measurements at 1 AU (Lugaz et al.2017b) and numerical simulations (Poedts, Pomoell, and Zuccarello 2016) indicate that slow CMEs may be able to drive shocks in part due to their large and long-lived expansion. It is unclear yet where these shocks form, as it depends on the relative decrease with radial distance of the ambient coronal and solar-wind plasma parameters, the solar-wind fast magnetosonic speed, and the CME (ejecta) speed including its expansion speed. The observed occurrence of ICME leading shocks at 1 AU varies in phase with solar activity and is on average about 65% (Jian, Russell, and Luhmann, 2011; Jian et al.2018d). The magnetosonic Mach number of these shocks is generally 1.2 to 4. Kilpua et al. (2015) found little solar-cycle variation in these numbers, suggesting that the combination of ICME and solar-wind properties both change in such a way as to maintain this general range of shock strengths. Using Helios 1/2 data, Lai et al. (2012) studied the radial variation of the magnetic-field compression ratios of 50 quasi-perpendicular shocks from 0.3 to 1 AU, as a proxy for the radial variation of Mach number, finding that the Mach number of ICME shocks does not vary much with heliocentric distance.

The shocks driven by ICMEs often have large proton foreshocks, regardless of whether the shocks are quasi-perpendicular or quasi-parallel (Blanco-Cano et al.2016). The foreshocks can range in upstream extent to more than 0.1 AU, relatively greater than their planetary bow-shock counterparts. This may be due to the shocks forming close to the Sun that then energize particles for longer times as they propagate to 1 AU. MESSENGER, with its SEP-detection capability, provided opportunities to evaluate the radial dependence of peak SEP intensities, an indirect diagnostic of CME and ICME shocks. For a selected set of events, MESSENGER was lined up along the interplanetary spiral field with one of the distributed 1 AU spacecraft. Lario et al. (2013) studied the radial and longitudinal distribution of \(\approx 100\) keV SEP electrons using MESSENGER, STEREO, and ACE data. They found a radial falloff of the peak SEP intensity significantly greater than the \(R^{-3}\) dependence expected from theory. In general, the SEP part of the ICME greater consequences is also tied to the overall understanding of its onset and evolution.

5 ICME Radial Evolution: As Revealed by Numerical Simulations

The past decade has seen multiple developments with respect to numerical simulations of CMEs and ICMEs including the routine usage and further development of models appropriate for real-time forecasting, Sun-to-1 AU simulations capable of producing synthetic observations (EUV, coronagraphs, HI, and/or in situ), and coronal simulations of complex initiation mechanisms, with the results sometimes extended all the way to 1 AU.

Heliospheric MHD codes such as ENLIL (Odstrcil 2003; Odstrcil, Pizzo, and Arge 2005; Odstrcil and Pizzo 2009) are now used in real-time for space-weather forecasting but also as support for research, especially to visualize and follow ICME heliospheric propagation and determine which ICMEs may impact a planet or interplanetary spacecraft. ENLIL results feature a solar wind based on inner heliospheric boundary conditions at \(21.5~\mbox{R}_{\odot}\) from the semi-empirical Wang–Sheeley–Arge (WSA) model (Arge and Pizzo 2000) or at \(30~\mbox{R}_{\odot}\) from the MHD-Algorithm-outside-a-Sphere (MAS) coronal model with polytropic and thermodynamic versions (e.g. Lionello, Linker, and Mikic 2009; Lionello et al.2013). The WSA model uses solar-surface-field synoptic maps constructed from magnetograph observations, together with a modified potential-field source-surface (PFSS) coronal-field model, to provide a first-order description of the initial stream structure and interplanetary field. Aspects of the ICME transients are included by applying the so-called “cone model” in which a (typically conic-section shaped or spherical) high-pressure pulse with direction, location and speed based on coronagraph images of CMEs is introduced into the simulation at its inner boundary (e.g. Mays et al.2015). ENLIL has proven to be a broadly useful tool for interpreting widespread in-situ measurements of ICMEs. For example, the WSA-ENLIL+Cone model was used to produce simulated ICME “shocks” whose arrival times at different planetary and spacecraft locations were compared with in-situ observations (e.g. at Earth by Bain et al.2016, at Venus by Möstl et al.2018, at Mercury by Baker et al.2013, and at Mars by Lee et al.2017). Most often, models such as WSA-ENLIL are invoked to provide predictions or post-event assessments of plasma speed, interplanetary-field polarity, and the time of arrival at a certain location of a CME-initiated interplanetary shock (e.g. Mays et al.2015; Wold et al.2018; Riley et al.2018).

One of the main limitations of ENLIL cone-model simulations is the lack of the internal (driver) magnetic field of the ICME. Versions with flux-rope or spheromak CME ejecta inserted at the inner boundary are in development (Odstrcil, Savani, and Rouillard 2018). Including the internal magnetic field is essential to forecasting the geomagnetic potential of CMEs and capturing the physics of ICME evolution more accurately. Other heliospheric MHD codes for space-weather forecasting and research support are under development especially in Europe – the European heliospheric forecasting information asset (EUHFORIA: Pomoell and Poedts 2018; Poedts 2019; Verbeke, Pomoell, and Poedts 2019) – and in Japan – the Space-weather-forecast-Usable System Anchored by Numerical Operations and Observations–CME (SUSANOO: Shiota and Kataoka 2016). Similar to ENLIL, a major advantage of these numerical simulations starting at or above 0.1 AU is that they are inexpensive to run for multiple realizations of events in 3D. They are thus frequently used to complement remote observations in retrospective event analyses or in forecasting an observed CME’s effects at 1 AU. The information they can provide includes the longitudinal extent of the ICMEs, the magnetic connectivity of particular heliospheric locations to their shocks, and the related in-situ plasma parameters. They moreover give important insights into the nature and effects of their solar-wind interactions in transit (e.g. Prise et al.2015; Winslow et al.2016; Witasse et al.2017; Kilpua et al.2019). There are now ongoing efforts to make more routine simulations of complex CME/ICME simulations that are initiated at the solar surface for both research support and eventually space-weather forecasting (Borvikov et al.2017; Jin et al.2017). However, unraveling the detailed physics and phenomena of the CME–ICME relationship and its radial evolution, in realistic contexts, demands further developments of more physically complete simulations.

While still not in the regular-use domain, these state-of-the-art numerical simulations have been useful for retrospectively analyzing real cases where relatively complete observations and measurements related to CMEs and their evolution into ICMEs are available. Synthetic images derived from the simulations are often used to understand remote coronal and heliospheric observations (Lugaz et al., 2008, 2009; Manchester et al.2008; Odstrcil and Pizzo 2009; Shen et al.2018; Jin et al.2017). For example, they have been applied to the interpretation of the EUV waves (Chen, Fang, and Shibata 2005; Delannée et al.2008; Cohen et al.2009; Downs et al.2011). Vourlidas et al. (2013) used such simulations to analyze the nature of CMEs in terms of the presence/absence of a twisted magnetic-flux rope in white-light images. Lynch et al. (2010, 2016) and Lynch and Edmonson (2013) were able to reproduce the appearance of eruptions originating in helmet streamers and pseudostreamers. Lynch and Edmonson (2013) and Török et al. (2018) created detailed physical descriptions of the pre-existing magnetic topology before eruptions, including the coronal helmet streamers, pseudostreamers, and null points. A particular contribution to our understanding from the improving simulations concerns the eruptions of multiple CMEs from different active regions, referred to as “sympathetic” CMEs, which are ubiquitous during active times of the solar cycle (Török et al.2011; Lynch and Edmonson 2013). While calculations at this level of sophistication that include the 1 AU consequences for validation against in-situ observations remain a challenge, they are seeing increasing applications. Al-Haddad et al. (2019) compared the properties of ICMEs hypothetically observed by multiple spacecraft assuming two different simulated CME magnetic morphologies. They learned that it was difficult to infer the initial coronal structure from the in-situ signatures. Reinard, Lynch, and Mulligan (2012) used simulations and in-situ measurements to interpret the ion composition within the structures of ICMEs toward better mapping back to their solar origins.

Lastly, simulations, often involving complex initiation mechanisms, are at the center of current efforts to better understand the physical processes that occur during CME propagation into the heliosphere. Manchester et al. (2014) followed up on the study of observationally inferred CME erosion using in-situ measurements (Ruffenach et al.2012; Lavraud et al.2014) with a Sun-to-1 AU numerical simulation. They found that the erosion due to magnetic reconnection between the ICME and the interplanetary magnetic fields occurred, but was relatively limited. However, there was extensive magnetic-field reconfiguration as the simulated CME/ICME propagated from the Sun to 1 AU (see also Manchester, van der Holst, and Lavraud 2014). A similar, complex reconfiguration and distortion of the original coronal flux rope was evident in another Sun-to-Earth simulation with a different code and initiation mechanism (Török et al.2018). That study, moreover, found that complexity could result in highly different model-based magnetic-field temporal profiles determined for locations only 10 – 15° apart, as suggested by the earlier work of Riley et al. (2004). A recent analysis of multi-spacecraft measurements separated by \(\approx 0.7\)° confirmed that significant differences of the magnetic fields of ICMEs at these separations are possible (Lugaz et al.2018). The complex nature of the magnetic field inside CMEs, and its implications for ICMEs, was similarly highlighted in a simulation by Savani et al. (2013). Simulations are also particularly useful for investigating the physical processes occurring during CME–CME or CME–ambient structure interactions. As examples, simulations have been run to understand the evolution of ICME shocks propagating inside previous eruptions (Schmidt and Cargill 2004; Lugaz, Manchester, and Gombosi 2005; Mao et al.2017), super-elastic collisions between CMEs (Shen et al., 2012, 2013), and the previously discussed magnetic reconnection expected to occur between erupted structures and their surroundings, as well as within the structures themselves (Lionello et al.2013; Lugaz et al.2013).

The ultimate goal of CME/ICME simulations is the greater physical understanding of their variety of origins and outcomes. A vision for the future continues to be the ability to simulate the full, coupled space-weather chain of events based on observations of the Sun and in-situ validations (see Figure 8). However, earlier success, and the insights it brings, is likely to come by focusing on the simplest cases first. Perhaps these involve the isolated eruption of large filament channels associated with decayed active regions. While simulations have successfully reproduced the images of a slow, streamer-blowout CME (Lynch et al.2016), the realistic propagation/evolution of the structure and its solar-wind interaction en route to \(\approx 1~\mbox{AU}\) represents a necessary and achievable next step. The current solar minimum should provide additional simulation-worthy cases, especially of the relatively simple streamer blowout CMEs (e.g. Vourlidas and Webb 2018) that may dominate space-weather conditions over the next three years or so.

Figure 8
figure 8

Simulation results of Török et al. (2018) (organized from a review by Manchester et al.2017) illustrating the potential of simulating a CME/ICME event from its solar source to a heliospheric observer site at 1 AU. Such detailed simulations require sufficient observations to both accurately launch the erupting coronal structure in realistic surroundings and have it propagate through realistic solar, wind conditions. (Image reproduced with permission from Manchester et al. (2017), copyright by the authors.)

Full size image

6 Inner Heliosphere ICME Evolution: Future Prospects

This overview has only touched on the substantial progress in observations and physical understanding of ICMEs achieved in the last decade. Going forward, the outstanding issues and challenges include determining how the evolution of the Heliophysics System Observatory will affect our ability to both reconstruct their Sun-to-1 AU behavior and interpret it. SOHO, STEREO, and now SDO observations continue to be regularly used in conjunction with in-situ observations as well as global coronal and heliospheric models to determine which solar and interplanetary events are related, and how they are related. An updated sketch of what an ICME includes, as we currently understand it, is shown in Figure 9. The flux-rope-like driver continues to be central to the process, although its origin is now considered, and scrutinized, in much more detail. Figure 9 includes some key details described in this brief overview that are missing in Figure 1. These include near-Sun evolution involving an early-phase pressure wave or shock that gives way to the driven shock observed in the heliosphere. Expansion is regarded as an essential process in the ICME evolution, with its contribution to shock formation being case-dependent. Self-similar expansion of the coronal entity is no longer regarded as the norm for interpreting ICMEs. Realistic, 3D models of the originating coronal and solar-wind structures push the state of the art in describing ambient conditions, while various descriptions of injected coronal structures – with increasingly detailed attributes – are being tested. The heliophysics community is at the threshold of being able to routinely fit models of CME structures seen in multi-perspective coronal images that become the ICMEs observed at 1 AU, with the realistic influences of the ambient medium the next frontier. The availability of both the \(\mbox{L}_{1}\) assets and STEREO data with their well-separated (relative to \(\mbox{L}_{1}\)) perspective continues to both motivate this work and provide the observational basis for its validation. The ultimate goal is sufficient understanding to reproduce, with physics-based global models, the observed multipoint ICME attributes based on solar observations.

Figure 9
figure 9

Sketch showing some of the updates to our picture of ICMEs as described in this overview, including \(\mathbf{i}\)) the spherical shock associated with the early evolution of major CMEs, including the EUV wave that may be a signature of its low coronal intersection; ii) the development of the longer-lived driven shock that may at first strengthen with distance from the Sun, and can be detected in situ at and beyond 1 AU; iii) the various behaviors that are inferred as different ICMEs evolve, including deflections from the initial CME direction near the Sun, changes associated with subsequent solar-wind structure interactions, and classical nearly self-similar expansion. Additional effects not shown here include rotation, erosion, and interactions (and sometimes mergers) with other ICMEs. Which of these occurs depends both on the properties and location of the ejecta, and the conditions in the ambient medium (corona and solar wind).

Full size image

Although MESSENGER and VEX are no longer in operation, MAVEN’s ongoing measurements of plasma, field, and energetic particles at Mars, and MSL’s RAD surface-radiation measurements, rely on SOHO/ACE/Wind (and eventually IMAP) \(\mbox{L}_{1}\) and STEREO observations for understanding the local effects of solar activity and ICME impacts. Future opportunities include various alignments (radial and Parker Spiral) of Earth, Mars, and STEREO-A for in-situ studies, and quadrature configurations providing imaging of CMEs destined for various targets. These will also help set the heliospheric context for upcoming BepiColombo observations, and they will provide similar support for Venus flybys of the Parker Solar Probe and Solar Orbiter missions. Finally, STEREO-A is in a position to test the \(\mbox{L}_{5}\) concept of Earth space-weather monitoring (e.g. Lavraud et al.2016), including ICME evolution from Sun to Earth, on which we may someday all depend. The heliospheric imagers onboard STEREO have allowed us to image ICMEs directly as they impact spacecraft in the inner heliosphere (MESSENGER, VEX, ACE/Wind) that make in-situ measurements. MESSENGER highlighted, 30 years after Helios, the importance of having in-situ measurements of ICMEs in the inner heliosphere. The Parker Solar Probe is taking in-situ measurements while they are still in the upper corona and visible in the LASCO-C3 field-of-view. Its observations will further advance our understanding by providing a more pristine view of “young” ICMEs only a few hours after their initiation. These measurements, especially those obtained in conjunction with other spacecraft, will further advance our understanding of the origins of the magnetic ejecta, the formation of the shock waves, and the development of the ICME sheaths. In addition to providing a second near-Sun probe, during its extended mission, Solar Orbiter will provide the first comprehensive images of ICMEs from a vantage point away from the Ecliptic, allowing further investigations of ICME radial evolution including deflections and interactions. Finally, with the increasing uses of magnetographs on missions, including Solar Orbiter, the next major leaps in real event simulations are imminent.