With the availability of data and computational technologies in the modern world, machine learning (ML) has emerged as a preferred methodology for data analysis and prediction. While ML holds great promise, the results from such models are not fully unreliable due to the challenges introduced by uncertainty. An ML model generates an optimal solution based on its training data. However, if the uncertainty in the data and the model parameters are not considered, such optimal solutions have a high risk of failure in actual world deployment. This paper surveys the different approaches used in ML to quantify uncertainty. The paper also exhibits the implications of quantifying uncertainty when using ML by performing two case studies with space physics in focus. The first case study consists of the classification of auroral images in predefined labels. In the second case study, the horizontal component of the perturbed magnetic field measured at the Earth's surface was predicted for the study of Geomagnetically Induced Currents (GICs) by training the model using time series data. In both cases, a Bayesian Neural Network (BNN) was trained to generate predictions, along with epistemic and aleatoric uncertainties. Finally, the pros and cons of both Gaussian Process Regression (GPR) models and Bayesian Deep Learning (DL) are weighed. The paper also provides recommendations for the models that need exploration, focusing on space weather prediction.
The geospace response to coronal mass ejections includes phenomena across many regions, from reconnection at the dayside and magnetotail, through the inner magnetosphere, to the ionosphere, and even to the ground. Phenomena occurring in each region are often connected to each other through the magnetic field, but that field undergoes dynamic changes during storms and substorms. Improving our understanding of the geospace response to storms requires a global picture that enables us to observe all the regions simultaneously with both spatial and temporal resolution. Using the Energetic Neutral Atom (ENA) imager on the Two Wide-Angle Imaging Neutral-Atom Spectrometers (TWINS) mission, a temperature map can be calculated to provide a global view of the magnetotail. These maps are combined with in situ measurements at geosynchronous orbit from GOES 13 and 15, auroral images from all sky imagers (ASIs), and ground magnetometer measurements to examine the global geospace response of a coronal mass ejection (CME) driven event on March 12th, 2012. Mesoscale features in the magnetotail are observed throughout the interval, including prior to the storm commencement and during the main phase, which has implications for the dominant processes that lead to pressure buildup in the inner magnetosphere. Auroral enhancements that can be associated with these magnetotail features through magnetosphere-ionosphere coupling are observed to appear only after global reconfigurations of the magnetic field.
We analyze data returned by the Magnetospheric Multiscale mission (MMS) constellation during a rapid (∼1.5 s) traversal of a flapping and reconnecting current sheet (CS) in the near-Earth magnetotail (X ∼−20 RE). The CS was highly tilted, with its normal pointing strongly duskward. Its extreme thinness was confirmed by a curvature analysis of the magnetic field lines. The event was associated with a guide field of 8% of the reconnecting components. From the pitch angle distributions of low-energy electrons we infer a crossing earthward of the X-line. Traveling practically normal to the CS, MMS encountered an ion diffusion region (IDR) in which was embedded an electron diffusion region (EDR). IDR signatures included breaking of the ion frozen-in condition in the presence of Hall B and E fields. EDR signatures included a strong out-of-plane current associated with a superAlfvénic electron jet, positive energy transfer, and a temperature anisotropy (Te∥ > Te⊥) which disappeared at the field reversal. Derived scale sizes normal to the CS are: ∼6.9 de (EDR) and ∼0.4 di (IDR; 40 and 100 km). We estimate the average dimensionless reconnection rate as 0.077 ± 0.050. The observations and inferences are supported by particle-in-cell (PIC) numerical simulations. We find very good agreement in the reconnection rates. We also discuss the effects of asymmetries in the density, temperature and magnetic field strength on the Hall fields and length of the outflow jets. The event is associated with a substorm onset which began 7 min after the MMS observations.
The NASA Two Wide-Angle Imaging Neutral Atom Spectrometers (TWINS) Mission of Opportunity is the first stereoscopic imaging mission of the magnetosphere. Each of two satellites hosts an energetic neutral atom (ENA) imager and a Lyman-alpha detector (LAD). The remote detection nature of these two instruments enables TWINS to make global observations of the magnetosphere. Such global measurements provide an excellent platform for the study of dawn-dusk asymmetries that appear in many characteristics of the magnetosphere. This work reviews the studies using TWINS data that have discussed dawn-dusk asymmetries over the past 7 years, expanding upon the review of the first 5 years of the mission by Goldstein and McComas , while focusing specifically on the analysis of the asymmetries.