Search for: All records

Abstract We develop an Imbalanced Regression Artificial Neural Network model for the Auroral electrojet index (IRANNA) to predict the SuperMAG SML index, addressing the heavily imbalanced distribution of the SML data set. The data set contains mostly quiet‐time values of lesser importance and very few strong‐to‐extreme values of interest, such as those associated with super substorms. Traditional prediction models, which minimize mean squared error uniformly across the whole data set, are often skewed by this imbalance, prioritizing the lower, quiet‐time values and consequently underestimating strong geomagnetic events. The IRANNA model addresses this issue by using a customized weighting scheme in the loss function, enabling it to predict strong‐to‐extreme events accurately for the first time. The model takes solar wind parameters as inputs and predicts the logarithm of the absolute SML values. It does not rely on past values of the SML index, differentiating it from other models that use historical data for prediction. The model has demonstrated its ability to predict the peak amplitudes of strong‐to‐extreme events across various statistical analyses, event studies, and virtual experiments. Despite this success, challenges remain, particularly during localized electrojet events and when upstream solar wind data propagation is unreliable. This study emphasizes the importance of using imbalanced regression techniques, especially in space physics, where data sets are inherently skewed. It also highlights the potential of the IRANNA model to provide valuable insights into the magnetosphere's response to solar wind driving, improving space weather forecasting and offering new tools for investigating magnetospheric dynamics. 

Abstract Whistler mode waves scatter energetic electrons, causing them to precipitate into the Earth's atmosphere. While the interactions between whistler mode waves and electrons are well understood, the global distribution of electron precipitation driven by whistler mode waves needs futher investigations. We present a two‐stage method, integrating neural networks and quasi‐linear theory, to simulate global electron precipitation driven by whistler mode waves. By applying this approach to the 17 March 2013 geomagnetic storm event, we reproduce the rapidly varying precipitation pattern over various phases of the storm. Then we validate our simulation results with POES/MetOp satellite observations. The precipitation pattern is consistent between simulations and observations, suggesting that most of the observed electron precipitation can be attributed to scattering by whistler mode waves. Our results indicate that chorus waves drive electron precipitation over the premidnight‐to‐afternoon sector during the storm main phase, with simulated peak energy fluxes of 20 erg/cm²/s and characteristic energies of 10–50 keV. During the recovery phase, plume hiss in the afternoon sector can have a comparable or stronger effect than chorus, with peak fluxes of ∼1 erg/cm²/s and characteristic energies between 10 and 200 keV. This study highlights the importance of integrating physics‐based and deep learning approaches to model the complex dynamics of electron precipitation driven by whistler mode waves. 

Abstract Electron density plays an important role in the study of wave propagation and is known to be associated with the index of refraction and radiation belt diffusion coefficients. The primary objective of our investigation is to explore the possibility of implementing an onboard signal processing algorithm to automatically obtain electron densities from the upper hybrid resonance traces of wave spectrograms for future missions. U‐Net, developed for biomedical image segmentation, has been adapted as our deep learning architecture with results being compared with those extracted from a more traditional semi‐automated method. As a product, electron densities and cyclotron frequencies for the entire DSX mission between 2019 and 2021 are acquired for further analysis and applications. Due to limited space measurements, a synthetic image generator based on data statistics and randomization is proposed as an initial step toward the development of a generative adversarial network in hopes of providing unlimited realistic data sources for advanced machine learning. 

Abstract Whistler mode waves in the plasmasphere and plumes drive significant losses of energetic electrons from the Earth's radiation belts into the upper atmosphere. In this study, we conducted a survey of amplitude‐dependent whistler wave properties and analyzed their associated background plasma conditions and electron fluxes in the plasmasphere and plumes. Our findings indicate that extremely large amplitude (>400 pT) whistler waves (a) tend to occur atL > 4 over the midnight‐dawn‐noon sectors and have small wave normal angles; (b) are more likely to occur during active geomagnetic conditions associated with higher fluxes of anisotropic electrons at 10 s keV energies; and (c) tend to occur at higher latitudes up to 20° with increasing amplitude. These results suggest that extremely large amplitude whistler waves in the plasmasphere and plumes could be generated locally by injected electrons during substorms and further amplified when propagating to higher latitudes. 

Abstract Whistler‐mode hiss waves are crucial to the dynamics of Earth's radiation belts, particularly in the scattering and loss of energetic electrons and forming the slot region between the inner and outer belts. The generation of hiss waves involves multiple potential mechanisms, which are under active research. Understanding the role of hiss waves in radiation belt dynamics and their generation mechanisms requires analyzing their temporal and spatial evolutions, especially for strong hiss waves. Therefore, we developed an Imbalanced Regressive Neural Network (IR‐NN) model for predicting hiss amplitudes. This model addresses the challenge posed by the data imbalance of the hiss data set, which consists of predominantly quiet‐time background samples and fewer but significant active‐time intense hiss samples. Notably, the IR‐NN hiss model excels in predicting strong hiss waves (>100pT). We investigate the temporal and spatial evolution of hiss wave during a geomagnetic storm on 24–27 October 2017. We show that hiss waves occur within the nominal plasmapause, and follow its dynamically evolving shape. They exhibit intensifications with 1 and 2 hr timescale similar to substorms but with a noticeable time delay. The intensifications begin near dawn and progress toward noon and afternoon. During the storm recovery phase, hiss intensifications may occur in the plume. Additionally, we observe no significant latitudinal dependence of the hiss waves within |MLAT| < 20°. In addition to describing the spatiotemporal evolution of hiss waves, this study highlights the importance of imbalanced regressive methods, given the prevalence of imbalanced data sets in space physics and other real‐world applications. 

Abstract The present study uncovers the fine structures of magnetosonic waves by investigating the EFW waveforms measured by Van Allen Probes. We show that each harmonic of the magnetosonic wave may consist of a series of elementary rising‐tone emissions, implying a nonlinear mechanism for the wave generation. By investigating an elementary rising‐tone magnetosonic wave that spans a wide frequency range, we show that the frequency sweep rate is likely proportional to the wave frequency. We studied compound rising‐tone magnetosonic waves, and found that they typically consist of multiple harmonics in the source region, and may gradually become continuous in frequency as they propagate away from source. Both elementary and compound rising‐tone magnetosonic waves last for ∼1 min which is close to the bounce period of the ring proton distribution, but their relation is not fully understood.