Project topics and materials
Get Complete Project Material File(s) Now! »
In-situ observations of the solar wind-magnetosphere coupling
The magnetosphere and the solar wind constitute a complex duet in permanent interaction at every scale. The variations of the solar wind modify the properties of the near-Earth regions, their boundaries, and generate small-scale physical processes such as magnetic reconnection that strongly affects the dynamics of this interaction. On the other hand, large-scale solar events such as Coronal Mass Ejections (CMEs), perturb the solar wind during their propagation in interplane-tary space and strongly affect the magnetosphere by the formation of geomagnetic storm that can have huge consequences on the human activity.
With the easy access to this specific region of the interplanetary medium for a spacecraft launched from Earth, numerous are the missions focused on the study of the solar wind and its relation with the magnetosphere whether they are solar wind monitors at the Lagrange point L1 (Wind and ACE in particular) or explorers of the different regions of the near-Earth environment (Cluster, Double Star, THEMIS and MMS to mention just a few of them).
The increasing number of these missions and the associated analysis of their in-situ data mea-surements led to the multiplication of case studies that increased our knowledge on the different physical processes at stake.
Nowadays, the accumulation of decades of spacecraft in-situ data measurement allows the elaboration of massive, global statistical studies of the different actors of the solar wind-magnetosphere interaction that use the data of several different missions at the same time. In particular, we can cite the studies that have been made on on the near-Earth regions (Zhang et al. [2019], Lavraud et al. [2004a] and references therein), their boundaries (Hasegawa [2012], Paschmann et al. [2018], Nˇemeˇcek et al. [2020] and references therein) and the physical processes at small (Lewis and Fuselier [2011], Hoshi et al. [2018], and references therein) and large scales (Kilpua et al. [2017], Richardson [2018], Chi et al. [2016] and references therein) that rule the interaction of the magne-tosphere with the solar wind.
As they concatenate an important number of samples, such statistical studies contribute to the elaboration of a global, statistical vision of the different physical processes that affect the Sun-Earth relation. Nevertheless, they often rely on the manual selection of events of interest in the streaming in-situ time series data provided by spacecraft. This, in addition to being time-consuming, is an ambiguous task, strongly linked to the interpretation of an external observer and poorly reproducible 3. This necessarily limits the information one can extract from the asso-ciated statistical studies. With the increasing number of spacecraft dedicated to the study of the near-Earth environment and the ever-growing amount of data provided by the totality of these spacecraft, the proportion of selected data will represent an even smaller proportion of the total accessible data, which spoil the potential of their overall consideration.
From now on, the elaboration of automatic event detection methods in streaming in-situ time series data provided by spacecraft appears as an interesting option to accelerate the collection of data and improve the reproducibility of statistical studies. For this purpose, manually set thresh-olds on the values of physical quantities appear as the most intuitive and fastest solution one can think of to improve the detection [Jelínek et al., 2012; Lepping et al., 2005]. Nevertheless, these methods are limited by the high variability of in-situ data and the manual setting of opti-mal thresholds is another time-consuming, ambiguous and hardly reproducible task, limited by the visual inspection of huge quantities of data. Moreover, these methods are often tested on the dataset on which they have been developed and there are generally no clues on how well they do on unknown sets of data and how easy it is to apply them to the data of different missions. Addi-tionally, the thresholds are often based on a reduced number of parameters and lose in efficiency when several physical features must be considered at the same time.
An interesting option we have to overcome these constraints then stands in using supervised machine learning algorithms. These algorithms, that have the ability to learn to perform a certain task after being trained on a given dataset, represent a promising tool to tackle already large and ever-growing bases of reliable data accumulated for decades, and their use in space physics is therefore progressing [Camporeale, 2019].
To what extent can these algorithms help us improve the automatic detection of the signatures in streaming in-situ data of the different physical processes that rule the interaction between the solar wind and the Earth magnetosphere ? How can they help constructing a global, statistically representative vision of these processes ?
The objectives of this thesis are then as follows:
• Investigate the potential of different machine learning algorithms in the optics of the fast, automatic and reproducible identification of the in-situ signatures of the different processes that intervene in the Sun-Earth interaction.
• Use these algorithms to perform a massive detection of the events of interest in the accu-mulated decades of in-situ data and generate some of the most exhaustive events catalogs.
• Exploit these catalogs to take a step further in the improvement of our global vision of the magnetosphere through the realization of massive, global statistical studies of the phenom-ena at stake.
Naturally, the question can be asked at every level of the interaction, and the application of machine learning algorithms thus has potential all along the chain of events from the Sun to the magnetosphere as this will be discussed in the next sections.
Large-scale solar events
In addition to the production of the solar wind, the physical processes that occur in the solar corona can generate large-scale solar events which, through the transport of important quantities of plasma and magnetic field, induce large-amplitude perturbations of the nominal solar wind with possible serious consequences regarding the dynamics of the near-Earth environment.
Among them, CMEs are the spectacular expulsion of large quantities of plasma and magnetic field in the interplanetary medium from the solar corona.
Suggested by the theoretical work of Chapman and Ferraro [1929] to explain geomagnetic dis-turbances in the Earth magnetosphere, their existence was confirmed by the ground observations of Hansen et al. [1971] and the on-board observations of Tousey [1973].
Produced in the solar corona, the ejecta propagates and expands in the interplanetary medium. This propagation of the ejecta, the so-called ICME, was first observed in by the measurmements of the interplanetary proton density and temperature by the Vela 3 spacecraft [Gosling et al., 1973] and linked to the production of CMEs by the combination of the data of 5 different spacecraft by Klein and Burlaga [1981].
The observation of a CME with the coronograph of the Solar and Heliospheric Observatory (SOHO) spacecraft along with the schematic representation of an ICME arriving at Earth orbit are shown on the two panels of Figure 1.5.
The propagation of the ejecta in the interplanetary medium is accompanied by a three-dimensional expansion. At 1 AU, the average estimated width of an ICME is in the order of 0.3 AU [Wang et al., 2005]. The transported magnetic field is commonly described as having helical field lines form-ing a so-called flux rope [Goldstein, 1983]. By definition, the amplitude of the magnetic field is stronger than the average IMF and the typical values of the magnetic field of ICMEs found at 1 AU is in the order of 10 nT [Kilpua et al., 2017]. Following the ejection at large velocities in the solar corona, the transported plasma is usually faster than the preceding nominal solar wind, at 1 AU, the average velocity found for ICMEs is in the orders of 450 km/s. When the velocity of the ejecta is fast enough, the ICME pushes the downstream solar wind at supersonic speeds and is thus drives a shock. The propagation of this shock wave is at the origin of a turbulent plasma region of low anisotropy between the shock and the main body of the ICME, the so-called sheath [Klein and Burlaga, 1981; Moissard et al., 2019].
Figure 1.5: CME observed by the SOHO spacecraft on the 27th of February 2000 (right) and schematic rep-resentation of an ICME (right) (Adapted from Zurbuchen and Richardson [2006]). These are the general average properties of ICMEs and these properties are also visible in solar wind observations during such events as shown in the Figure 1.6. If the different characteristics we just described define the commonly agreed in-situ signatures of these events, their identification is actually much more ambiguous and strongly related to the interpretation of an external observer [Shinde and Russell, 2003]. In Chapter 3, we will give a particular focus on how the application of machine learning highlights the difficulties inherent to the identification of ICMEs from in-situ data measurement.
The link between the CMEs and the geomagnetic disturbances recorded in the Earth magne-tosphere was confirmed by Wilson [1987] who noticed the particular importance of the southward component of the IMF on the triggering of geomagnetic storms. From then on, storms were also observed for northward IMF [Du et al., 2008] and the CMEs are considered as the most geoeffective solar events [Yermolaev et al., 2012] at the origin of the greatest part of geomagnetic storms [Echer et al., 2005]. Among the various hazards already caused by these events on the human activity, one can typically cite the Bastille Day event [Webber et al., 2002], one of the largest geomagnetic storm ever recorded in space, or the March 1989 geomagnetic storm that led to an electrical power blackout in the entire Quebec [Boteler, 2019].
Figure 1.6: Solar wind observation during an ICME from the WIND spacecraft located at the Lagrangian Point L1. The solid vertical lines delimitate the ICME while the dashed vertical line indicate the beginning of the sheath. From the top to the bottom are represented : the magnetic field amplitude and components, the proton density and the solar wind velocity.
Eventhough the geoeffectiveness is expected to be related to the large quantities of plasma and magnetic field transported by the ejecta [Turc et al., 2014], the physical properties of ICMEs that are the most likely to affect the magnetospheric activity are still under debate [Kilpua et al., 2017].
In the development of space weather, a global, statistically representative vision of CMEs would be the opportunity to better understand the nature of these events and how they interact with the magnetosphere and affect the human activity. For this purpose, the increasing number of solar wind oriented missions (SOHO, STEREO, WIND, ACE, SOlar Orbiter, Parker Solar Probe just to mention a few of them) led to the multiplication of the existing ICMEs catalogs and associated statistical studies of their different physical parameters [Lepping et al., 2006; Nieves-Chinchilla et al., 2018; Richardson and Cane, 2010]. Nevertheless, the lack of consensus on the typical in-situ signature of ICMEs associated to the fact these catalogs were elaborated after a manual selection of events resulted in incomplete, ambiguous and hardly reproducible lists which using masks the statistical vision we can have on such events.
In this context, elaborating automatic detection methods would allow the rapid and repro-ducible collection of such events for their further statistical analysis. Moreover, the analysis of the events detected by one of these methods could bring interesting information on how visual identification is made and how we interpret in-situ data measurements.
We will focus on those questions in the Chapter 3 that will entirely be dedicated to the auto-matic detection of ICMEs.
If CMEs are known so, they are far from being alone in the zoology of the large-scale solar events produced in the solar corona which transport of plasma and magnetic field strongly affects the Earth magnetosphere. Among this zoology, we can particularly cite the Corotating Interaction Regions (CIRs), the interaction of a stream of high speed solar wind emanating from coronal holes with the preceding slower nominal solar wind (Richardson [2018] and references therein), and the interplanetary shocks, direct consequence of the propagation of CMEs and CIRs in the interplan-etary medium and responsible for the acceleration of particles to very high energies (Oliveira and Samsonov [2018] and references therein).
The magnetopause, boundary between the solar wind and the mag-netosphere
MHD discontinuities
The different regions of the near-Earth environment are characterised by different physical prop-erties. Consequently, the boundaries that delimit the three regions, the magnetopause and the bow shock, are discontinuities that evolve with the interaction of the plasma of the different me-dias.
In the frame of ideal Magnetohydrodynamics (MHD), when the plasma is considered as per-fectly conducting, these discontinuities can be described by the Rankine-Hugoniot equations Where ‰ is the density, P is the thermal and kinetic pressure, [X] the jump of a parameter X across the discontinuity, Xn and Xt denotes respectively the normal and the tangential components of X .
When [Vn ] 6˘0, this is particularly what happens at the interface between the magnetosheath and the solar wind, the discontinuity is a shock.
Otherwise, we can distinguish three different configurations schematically represented in the Figure 1.7 :
• If Vn ˘ 0 and Bn 6˘0, all the physical parameters but the density are continuous. No mass flow across the discontinuity is allowed and the jump in density is compensated by a jump in thermal pressure that rapidly disperse this so-called contact discontinuity.
• If Vn ˘ 0 and Bn ˘ 0, the discontinuity is said tangential (TD). The flow and the magnetic field are tangential to the discontinuity, the total pressure on the two sides balance and no mass or magnetic flux crossing is allowed. This is what happens at the magnetopause when no penetration of solar wind plasma is allowed.
• If Vn 6˘0, the discontinuity is said rotational (RD). Mass flow crossing is allowed, the tan-gential velocity and magnetic field rotate but keep their magnitudes constant and equal to the Alfven velocity across the discontinuity. This is for instance what happens when the IMF reconnects with the geomagnetic field as this will be detailed in the next section.
Location and shape of the magnetopause
At first sight, magnetopause can locally be approximated by a TD and prevents any transport of the shocked solar wind into the magnetosphere. This boundary can be defined by the surface where the magnetosphere and the solar wind total pressure balance each other. In the solar wind, the total pressure can easily be approximated by the lone dynamic pressure Pd yn ˘ ‰sw Vsw2 because of the weak magnitude of the IMF and thermal pressure. In the magnetosphere, the thermal and dynamic pressures can be neglected in comparison to the magnetic pressure.
Where BE is the magnetic field measured at the surface of the Earth and • accounts for the deviation of the magnetic field from its dipolar value and the field generated at the boundary sur-face. Assuming • ˘ 2, BE ˘ 3.1 £ 104 nT, and Pd yn ˘ 2 nPa, we obtain a typical value expected for the stand-off distance under standard solar wind conditions: r0 ˘ 9.9 Re.
An expression of the position and shape of the magnetopause can then be obtained by solving the equation (1.6). This is what was done analytically by Spreiter and Briggs [1962] and numerically by Sotirelis and Meng [1999].
A typical representation in the meridional plane of the magnetopause obtained from pressure balance is shown in the Figure 1.8. In addition to the stand-off distance, the level of flaring that describes the expansion of the surface in both equatorial and meridional planes is a key factor that characterises the position and shape of the magnetopause for changing solar wind conditions.
By solving the equation 1.6, Spreiter and Briggs [1962] noticed that the topological change of the geomagnetic field line at the polar cusps induced discontinuities between the dayside and the nightside magnetopause in these regions. These discontinuities are represented by the singular points on the magnetopause in the two hemispheres in the Figure 1.8. In practice, the magne-topause can be continuously extended in these regions through the introduction of an indentation that consider the geometry of the polar cusps with varying solar wind and seasonal conditions.
Following this theoretical definition, the first observation of the magnetopause was made with the measurements of Explorer 12 by Cahill and Amazeen [1963]. From then on, the accumulation of the missions that came across the different boundaries of the near-Earth environment allowed the multiplication of the studies focused on the magnetopause, whether they concern its location and shape (Nˇemeˇcek et al. [2020] and references therein) or its global dynamics (Hasegawa [2012]; Paschmann et al. [2018] and references therein). The collection of several observed magnetopause crossings allowed the establishment of empirical analytical models of the magnetopause shape and location that kept improving with the evidences of the influences of the different solar wind and seasonal parameters (Fairfield [1971]; Jelínek et al. [2012]; Lin et al. [2010]; Liu et al. [2015]; Shue et al. [1997] just to mention a few).
These observations also proved that the magnetopause is the theater of small-scale plasma processes that contribute to the dynamics of the boundary. Among them, magnetic reconnection fundamentally affects the location and shape of the magnetopause through the convection of the geomagnetic field lines it rearranges. Naturally, the evidence of this phenomenon questions the first definition of the magnetopause we gave. How do this process affect the position and shape of the magnetopause ? How can we consider it in the frame of an analytical model fitted from in-situ data?
The generation and the characteristics of these processes is strongly dependent on the asso-ciated upstream solar wind conditions. Consequently, their effects on the magnetopause location and shape are seen through the variations of the boundary with changing solar wind and seasonal conditions.
With the important number of studies dedicated to the subject, the influence of some of these parameters has been showed for long: the dynamic pressure pushes the magnetopause earthward when increasing and the IMF Bz component increases the azimuthal flaring while reducing the equatorial one when negatively decreasing. The importance of some other parameters, the two other components of the IMF Bx and By for instance, is however unclear and still under debate. Additionally, the previous existing studies are limited in the night side and there is no indication if what we know about the magnetopause holds in the far night side where the identification of the magnetopause becomes even more ambiguous. Last but not least, the existence of magnetic reconnection fundamentally affects the topography of the polar cusp and blurs the nature of the magnetopause in this region. There is thus no clue about the reality of the theoretically predicted indentation.
In the wake of the existing studies, answering these open questions requires the collection of as many magnetopause crossings as possible.
A typical in-situ signature of a magnetopause crossing by the THEMIS E spacecraft is represented in the Figure 1.9. At first, the high density above 20 cm¡3 and the omnidirectional differen-tial energy ion fluxes indicate the spacecraft is in the magnetosheath, as the velocity, shown on the third panel, is low, the crossing happens near the stagnation point. Past 12 : 00, one can notice a drop in density followed by a jump in Bz , the spacecraft has crossed the magnetopause and is now in the magnetosphere. The interpretation of the high velocity peaks highlighted with the green intervals will be detailed in the next section.
In practice, the in-situ measurement of magnetopause crossings are not as clear in every re-gion of the near-Earth environment and for every upstream solar wind condition and the iden-tification of such events much less obvious 4. The motion of the magnetopause with changing conditions result in partial crossings and the nature of the data measured by spacecraft with polar orbit is quite different from the nature of the data measured on an equatorial orbit. Additionally, the manual selection of such events is necessarily ambiguous and time-consuming. The automa-tion of this task would then be an important improvement in the elaboration of statistical studies of the different properties of the magnetopause.
Because of the drawbacks of methods based on manually-set thresholds, one can see the po-tential of machine learning algorithms in the realisation of this task. The application of such algo-rithms in the frame of the automatic detection of the near-Earth regions and boundaries will be discussed in the Chapter 4 and the magnetopause crossings detected with these methods will be exploited through the statistical study of the magnetopause shape and location in the Chapter 5.
Small-scale physical processes of the near-Earth environment
Because the solar wind and the magnetosphere are two plasmas of different nature, their inter-action is likely to create small-scale physical processes that strongly affects the dynamics of the system Among them, we can for example mention Kelvin-Helmholtz instability that results in the propagation of surface waves along the magnetopause (Kivelson and Zu-Yin [1984] and references therein) or magnetic reconnection that occurs when two non-parallel field lines are merged and topologically rearranged.
The latter was evidenced as the dominant process when it comes to the transfer of momentum between the solar wind and the magnetosphere [Sibeck et al., 1999]. For this reason, the part of this thesis dedicated to the small-scale physical processes of the near-Earth environment focuses on magnetic reconnection.
Magnetic reconnection
Magnetic reconnection is likely to occur when two conductive plasma with non-parallel magnetic field interact with each other. The interacting field lines are merged resulting in a topological re-arrangement of the system, often characterised by the conversion of magnetic energy into kinetic and thermal energy.
The term magnetic reconnection was first used by Dungey [1953] who showed that a break-down of the frozen-in law could result in the rearrangement of the field lines connectivity and associated particle acceleration. From then on, the decades of studies dedicated to the compre-hension of the process evidenced magnetic reconnection as a key factor of a wide range of phe-nomena. For instance, it plays a fundamental role in the formation of geomagnetic storms and au-roras, in the relativistic jets emitted by active galactic nuclei or sawtoothing oscillations observed in tokamak fusion plasmas (Yamada et al. [2010] and references therein). Magnetic reconnection is also believed to be the main actor at the origin of the formation or solar flares and CMEs that will constitute the main topic of the Chapter 3 [Shibata et al., 1995].
A schematic representation of magnetic reconnection is shown in the Figure 1.10. The incom-ing plasma flows are designated as the inflows while the flows of accelerated particles are denom-inated the outflows. When the frozen-in law is respected, the boundary between the two media is closed and no transfer of mass and momentum between the two plasmas is allowed. This is the case we described in the previous section when the plasma of the magnetosheath flows around the magnetopause.
With the violation of the frozen-in condition, the ions and the electrons decouple from the magnetic field in the proximity of the reconnection site. These regions of demagnetization are often called the diffusion regions and usually have a thickness in the order of the particles inertial length or thermal larmor radius, depending on the amplitude of the guide field. The Ion Diffusion Region (IDR) and the Electron Diffusion Region (EDR) are indicated in the Figure 1.10 with the grey and the blue rectangles respectively. The thick black lines of Figure 1.10 represent the field lines that are just being reconnected. As these lines separate the regions of different magnetic topology, they are called the separatrices. Because of the X shape they adopt, the point in the center of interest where they intersect is called the X-point. This is the point where the field lines are reconnected and from which they are convected with the outflow. In a 3D configuration, the counterpart of this reconnection site is called the X-line.
When all the physical parameters of the two interacting plasmas are equal but the orientation of their magnetic field, the reconnection is said to be symmetric. This well approximates what hap-pens in the magnetotail as this will be described in the next subsection. Otherwise, reconnection is said to be asymmetric. This is especially what happens at the magnetopause where the dense and cold magnetosheath interacts with the hot, tenuous magnetosphere.
Figure 1.10: Schematic representation of magnetic reconnection, the black lines represent the magnetic field lines of the two media, the thick black lines are the separatrix, the black arrows indicate the plasma inflow and outflow, the grey rectangle represents the IDR and the blue rectangle represents the EDR.
Magnetic reconnection at the Earth magnetopause
Reconnection of the IMF with the geomagnetic field was first detailed by Dungey [1961] and rep-resented in the Figure 1.11 in the case of a southward IMF.
The interplanetary field lines, represented in blue, and the geomagnetic field lines, represented in green, reconnect in the dayside of the magnetosphere at the reconnection site indicated by the grey rectangle (step 1 of the Figure). It is worth noting that reconnection in this region is asym-metric. The reconnected field lines (represented in red), open on one side and attached to the Earth pole on the other side, are convected tailward 5 by the flow of solar wind and pile up in the magnetotail (step 2 of the Figure). As the convected field lines point sunward in the north-ern hemisphere, and antisunward in the southern hemisphere, they reconnect in this region (step 3 ), interrupting the accumulation of flux in the process. On the nightward side of this symmetric reconnection site, which location is represented by the second grey rectangle, a bubble of plasma, known as a plasmoid, is expelled in the interplanetary medium. On the other side, the reconnected field lines are attached to the Earth and convected earthward carrying energetic accelerated particles which precipitation is at the origin of the formation of auroras (step 4 ). These closed field lines are then brought back to the dayside where they can reconnect again with the interplanetary field lines ensuring the continuity of the so-called Dungey cycle (step 5 ).
Following the description of Dungey [1961], when the IMF is northward and in a null dipole tilt condition, the interplanetary field lines and the closed geomagnetic field lines are parallel around the equatorial plane, indicating the absence of reconnection in this region. With the solar wind flowing around the magnetosphere, the interplanetary field lines are draped around the magne-topause and are likely to reconnect at high latitude where they are quasi anti-parallel to the geo-magnetic field lines. The sunward convection of the newly reconnected field lines that appears in this case is opposed the flow of the solar wind.
In both cases, reconnection fundamentally affects the position and shape of the magnetopause by eroding the magnetosphere in a direction that depends on the IMF orientation, on the dayside equatorial plane when it is southward and at high-latitudes when it is northward [Aubry et al., 1970]. In the previous observational studies of the magnetopause, the effects of this erosion are considered through the study of the influence of the lone IMF Bz component that appears to be the component with the greatest impact on the magnetic topology of the magnetosphere. How-ever, this consideration is reductive regarding the effect of the two other components of the IMF, Bx and B y , on magnetic reconnection, which has been evidenced for long [Gonzalez and Gonza-lez, 1980; Russell and Atkinson, 1973]. Consequently, the influence of these two other components of the IMF is still unclear and open to further investigations. This open question will be one of the main central topic of the chapter 5.
By reconfiguring the magnetic topology of the magnetosphere, reconnection modifies the na-ture of the near-cusp magnetopause. Indeed, the convection of the newly reconnected field lines creates boundaries that separate the plasma in the polar cusp from the magnetosheath called the cusp external boundaries [Lavraud et al., 2004b] by opposition with the so-called cusp inner boundaries that separate the cusp exterior from the magnetosphere. Without reconnection, the latter is the logical continuous extension between the dayside and the nightside magnetopauses. In the sense of reconnection, one of these boundaries actually corresponds to the separatrix of the geomagnetic and the interplanetary field liness. Consequently, the former appears as a more appropriate continuous extension of the magnetopause in the near-cusp region in the optics of reconnection that occurs whatever the orientation of the IMF might be. Although observed by various missions [Lavraud et al., 2004a; Zhou and Russell, 1997], the topology of this boundary for various solar wind conditions is still unclear and there are no existing clues on its actual indenta-tion. We will come back on this still-open question in the Chapter 5.
The first in-situ evidence of magnetic reconnection was brought by Sonnerup and Cahill Jr. [1967] who observed non-zero normal magnetic field component at the crossing of the magne-topause by Explorer 12, indicating an interface between the magnetosheath and the geomagnetic fields that was not limited to the lone tangential discontinuity. This first observation was followed by the first evidence of accelerated plasma at the magnetopause found by Paschmann et al. [1979] using the ISEE satellites and by the evidence of reconnection in the magnetotail observed in IMP data by Hones Jr. et al. [1976] that were consistent with the predictions of Dungey. From then on, the evidences of reconnection occurring at the magnetopause multiplied with the accumulation of missions. In particular, we can cite Cluster, THEMIS and Double Star, some of the missions that will focus our attention in this thesis. Nowadays, the technological advances allow an in-situ measurement of the plasma properties with an even finer time resolution permitting a deeper in-vestigation of the complexity of magnetic reconnection. For instance, the high resolution of the measurements of the recently launched MMS allowed an observational insight on the EDRs for the first time.
For now, the most solid evidence we can collect happens when a spacecraft goes through a reconnection outflow during the crossing of the magnetopause 6. Because of reconnection, the plasma of the outflow is accelerated, a spacecraft going through this region would then see a so-called plasma jet faster than the surrounding magnetosheath flow with a peaking component roughly corresponding to the component of the magnetic field that reverses during the crossing. This is especially the case for the jets represented by the green rectangles in the Figure 1.9. Here, the 5 jets we identified are oriented in the ¯Z direction indicating a spacecraft actually located north of the X-line, when the observed jets peak in both ¯Z and ¡Z directions, we usually talk about reversal jets that indicate a passage in the close proximity of the X-line by the spacecraft.
40 years of in-situ observation allowed an intensive study of the different aspects of magnetic reconnection that completed the theoretical and the numerical works dedicated to the subject. If these studies already confirmed a wide range of properties of magnetic reconnection, the clear influence of the conditions in the inflow regions [Cassak and Shay, 2007], the role of the different IMF components (Lavraud et al. [2005], Sonnerup [1974]) or its suppression by diamagnetic ef-fects [Swisdak et al., 2003] just to mention a few, the secrets of this process are far from all being unlocked. Among the remaining unknowns, one can especially cite the conditions that initiate the process, the structure of the different diffusion regions or the parametric dependence of the X-line location.
The latter has been under debate since the very first premises of studies dedicated to magne-topause magnetic reconnection. The question to know if reconnection occurs where the magnetic field of both sides are anti-parallel (anti-parallel reconnection [Crooker, 1979]) or if only a compo-nent of the IMF actually reconnects (component reconnection [Sonnerup, 1974]) has been under debate for years. De facto, both scenarios have been observed in spacecraft data and the question is then not to know which one prevails but how does the actual location of reconnection line, that should probably be a combination of these two configurations, varies with changing solar wind and seasonal conditions. Using Polar observations, Trattner et al. [2007] inferred that reconnec-tion occurs where the shear angle between the magnetosheath and the geomagnetic field lines is maximized and developed the so-called maximum shear angle model. Although the the capac-ity of this model to predict the local orientation of the X-line has been indicated in an important number of observational studies (Cassak and Fuselier [2016] and references therein), one does still not know how to link this empirical predictions to the different parameters we believe to af-fect reconnection dynamics such as the density, the field amplitude or bulk velocity jumps across the magnetopause [Cassak and Shay, 2007]. The numerical and theoretical investigations of the X-line orientation also led to the development of numerous models among which we can cite the maximisation of the outflow speed by Swisdak and Drake [2007], the bisection between the mag-netospheric and the magnetosheath fields by Aunai et al. [2016] or the maximisation of the current density of the magnetopause by Gonzalez and Mozer [1974]. Nevertheless, the comparison of all of these models, including the maximum shear angle model, to observational data has only been done through the investigation of the local orientation of the X-line [Souza et al., 2017] or on a small number of events [Trattner et al., 2012]. We then lack both of a global vision on how the re-connection sites extend over the whole magnetopause and of a comparison of these models with an important number of samples.
Table of contents :
1 Introduction
1.1 Solar wind and the near-Earth environment
1.2 Large-scale solar events
1.3 The magnetopause, boundary between the solar wind and the magnetosphere
1.4 Small-scale physical processes of the near-Earth environment
1.5 Summary
1.6 Bibliography
2 Machine learning as an automatic selection tool
2.1 Introduction
2.2 Different types of supervised classification
2.3 Evaluating the performances of an algorithm
2.4 ImplementingMachine Learning
2.5 Machine Learning in space physics
2.6 Bibliography
3 An example of ambiguously labeled problem: automatic detection of ICMEs
3.1 Introduction
3.2 Interplanetary CoronalMass Ejections
3.3 Data
3.4 Algorithm
3.5 Results
3.6 Robustness
3.7 Global quality of the prediction
3.8 Conclusion
3.9 Bibliography
4 Automatic classification of the three near-Earth regions
4.1 Introduction
4.2 Data
4.3 Labeling THEMIS data
4.4 Algorithm selection
4.5 Algorithm performance
4.6 Adaptability of the model: from a mission to the other
4.7 Comparison withmanually set thresholds
4.8 Massive detection of boundary crossings
4.9 Conclusion
4.10 Bibliography
5 Statistical analysis of the magnetopause shape and location
5.1 Introduction
5.2 A brief insight on themagnetopause shape and location models
5.3 Statistical analysis of the magnetopause crossing
5.4 Fitting a new magnetopausemodel
5.5 Nature of the near-cusp magnetopause
5.6 Conclusion
5.7 Bibliography
6 Automatic detection of magnetopause plasma flow
6.1 Introduction
6.2 Construction of the dataset
6.3 Jet detection pipeline
6.4 Method performance
6.5 Adaptability of the method
6.6 Massive detection of magnetopause plasma jets
6.7 Conclusion
6.8 Bibliography
7 Conclusions and prespectives
7.1 Overview
7.2 Application of supervised machine learning algorithms
7.3 Position and shape of the magnetopause
7.4 Potential of machine learning algorithms and larger perspectives
A Coordinates systems I
A.1 Geocentric Solar Ecliptic (GSE)
A.2 Geocentric SolarMagnetospheric (GSM)
A.3 LocalMagnetopause Normal (LMN)
A.4 Magnetic Local Time (MLT)
B Additional prediction examples III
B.1 ICMEs
B.2 Near-Earth regions
B.3 Cusp external boundary
B.4 Reconnection jets
C List of Acronyms XV
