Large geomagnetically induced currents (GICs) pose a risk to ground based infrastructure such as power networks. Large GICs may be induced when the rate of change of the ground magnetic field is significantly elevated. We assess the ability of three different machine learning model architectures to process the time history of the incoming solar wind and provide a probabilistic forecast as to whether the rate of change of the ground magnetic field will exceed specific high thresholds at a location in the UK. The three models tested represent feed forward, convolutional and recurrent neural networks. We find all three models are reliable and skillful, with Brier skill scores, receiver‐operating characteristic scores and precision‐recall scores of approximately 0.25, 0.95 and 0.45, respectively. When evaluated during two example magnetospheric storms we find that all scores increase significantly, indicating that the models work better during active intervals. The models perform excellently through the majority of the storms, however they do not fully capture the ground response around the initial sudden commencements. We attribute this to the use of propagated solar wind data not allowing the models notice to forecast impulsive phenomenon. Increasing the volume of solar wind data provided to the models does not produce appreciable increases in model performance, possibly due to the fixed model structures and limited training data. However, increasing the horizon of the forecast from 30 min to 3 h increases the performance of the models, presumably as the models need not be as precise about timing.

Using a particle-in-cell code, we study the diffusive response of electrons due to wave-particle interactions with whistler-mode waves. The relatively simple configuration of field-aligned waves in a cold plasma is used in order to benchmark our novel method, and to compare with previous works that used a different modelling technique. In this boundary-value problem, incoherent whistler-mode waves are excited at the domain boundary, and then propagate through the ambient plasma. Electron diffusion characteristics are directly extracted from particle data across all available energy and pitch-angle space. The 'nature' of the diffusive response is itself a function of energy and pitch-angle, such that the rate of diffusion is not always constant in time. However, after an initial transient phase, the rate of diffusion tends to a constant, in a manner that is consistent with the assumptions of quasilinear diffusion theory. This work establishes a framework for future investigations on the nature of diffusion due to whistler-mode wave-particle interactions, using particle-in-cell numerical codes with driven waves as boundary value problems.Plain Language Summary 'Whistler-mode' plasma waves interact with electrons in the Earth's outer radiation belts. This wave-particle interaction plays a significant role in both electron acceleration, and in the loss of electrons to the atmosphere via 'pitch angle scattering'. Such processes are typically modelled using numerical diffusion codes, with electron diffusion coefficients that characterize the nature and the strength of the wave-particle interaction. These diffusion coefficients are calculated using a mixture of long-established theory and input parameters taken from data and/or empirical models. We present a novel method for the direct extraction of characteristics of the electron diffusion from particle-in-cell numerical experiments. Our results demonstrate that the rate of diffusion can be time-dependent at early times, but then tends to constant values in a manner that is consistent with quasilinear theory.

Electron precipitation is a key component linking the ionosphere and the magnetosphere. Electrons in the magnetosphere-ionosphere (MI) system carry current, transport energy, and precipitate (i.e., follow magnetic field lines from the magnetosphere to the ionosphere) to collide with the neutral atmosphere thereby driving changes in the electrical conductivity tensor. This tensor is central to the three-dimensional electrical current circuit that flows over vast distances between the magnetosphere and the ionosphere. Indeed, particle precipitation is a key input to all global circulation models (GCMs) such as the Global Ionosphere Thermosphere Model (GITM) (Ridley et al., 2006), the Thermosphere Ionosphere Electrodynamics General Circulation Model (TIE-GCM) (Roble et al., 1988), and the Whole Atmosphere Model-Ionosphere

We present two new solar wind origin classification schemes developed independently using unsupervised machine learning. The first scheme aims to classify solar wind into three types: coronal-hole wind, streamer-belt wind, and 'unclassified' which does not fit into either of the previous two categories. The second scheme independently derives three clusters from the data; the coronal-hole and streamer-belt winds, and a differing unclassified cluster. The classification schemes are created using non-evolving solar wind parameters, such as ion charge states and composition, measured during the three Ulysses fast latitude scans. The schemes are subsequently applied to the Ulysses and the Advanced Compositional Explorer (ACE) datasets. The first scheme is based on oxygen charge state ratio and proton specific entropy. The second uses these data, as well as the carbon charge state ratio, the alpha-to-proton ratio, the iron-to-oxygen ratio, and the mean iron charge state. Thus, the classification schemes are grounded in the properties of the solar source regions. Furthermore, the techniques used are selected specifically to reduce the introduction of subjective biases into the schemes. We demonstrate significant best case disparities (minimum ≈8%, maximum ≈22%) with the traditional fast and slow solar wind determined using speed thresholds. By comparing the results between the in-(ACE) and out-of-ecliptic (Ulysses) data, we find morphological differences in the structure of coronal-hole wind. Our results show how a data-driven approach to the classification of solar wind origins can yield results which differ from those obtained using other methods. As such, the results form an important part of the information required to validate how well current understanding of solar origins and the solar wind match with the data we have.

Solar wind electron velocity distributions at 1 au consist of a thermal “core” population and two suprathermal populations: “halo” and “strahl”. The core and halo are quasi-isotropic, whereas the strahl typically travels radially outwards along the parallel or anti-parallel direction with respect to the interplanetary magnetic field. Using Cluster-PEACE data, we analyse energy and pitch angle distributions and use machine learning techniques to provide robust classifications of these solar wind populations. Initially, we used unsupervised algorithms to classify halo and strahl differential energy flux distributions to allow us to calculate relative number densities, which are of the same order as previous results. Subsequently, we applied unsupervised algorithms to phase space density distributions over ten years to study the variation of halo and strahl breakpoint energies with solar wind parameters. In our statistical study, we find both halo and strahl suprathermal breakpoint energies display a significant increase with core temperature, with the halo exhibiting a more positive correlation than the strahl. We conclude low energy strahl electrons are scattering into the core at perpendicular pitch angles. This increases the number of Coulomb collisions and extends the perpendicular core population to higher energies, resulting in a larger difference between halo and strahl breakpoint energies at higher core temperatures. Statistically, the locations of both suprathermal breakpoint energies decrease with increasing solar wind speed. In the case of halo breakpoint energy, we observe two distinct profiles above and below 500 km s−1. We relate this to the difference in origin of fast and slow solar wind.