20 October 2012 in Journal of Geophysical Research
28 November 2018 in Space Weather
In this study, we present a method that combines a Long Short‐Term Memory (LSTM) recurrent neural network with a Gaussian process (GP) model to provide up to 6‐hr‐ahead probabilistic forecasts of the Dst geomagnetic index. The proposed approach brings together the sequence modeling capabilities of a recurrent neural network with the error bars and confidence bounds provided by a GP. Our model is trained using the hourly OMNI and Global Positioning System (GPS) databases, both of which are publicly available. We first develop a LSTM network to get a single‐point prediction of Dst. This model yields great accuracy in forecasting the Dst index from 1 to 6 hr ahead, with a correlation coefficient always higher than 0.873 and a root‐mean‐square error lower than 9.86. However, even if global metrics show excellent performance, it remains poor in predicting intense storms (Dst < −250 nT) 6 hr in advance. To improve it and to obtain probabilistic forecasts, we combine the LSTM model obtained with a GP and evaluate the hybrid predictor using the receiver operating characteristic curve and the reliability diagram. We conclude that this hybrid methodology provides improvements in the forecast of geomagnetic storms, from 1 to 6 hr ahead.
5 June 2018 in Journal of Geophysical Research: Space Physics
We present a global model of plasmaspheric hiss, using data from eight satellites, extending the coverage and improving the statistics of existing models. We use geomagnetic activity dependent templates to separate plasmaspheric hiss from chorus. In the region 22‐14 MLT the boundary between plasmaspheric hiss and chorus moves to lower L∗ values with increasing geomagnetic activity. The average wave intensity of plasmaspheric hiss is largest on the dayside and increases with increasing geomagnetic activity from midnight through dawn to dusk. Plasmaspheric hiss is most intense and spatially extended in the 200‐500 Hz frequency band during active conditions, 400 <AE< 750 nT, with an average intensity of 1128 pT2 in the region 05‐17 MLT from 1.5 <L∗< 3.5. In the pre‐noon sector, waves in the 100‐200 Hz frequency band peak near the magnetic equator and decrease in intensity with increasing magnetic latitude, inconsistent with a source from chorus outside the plasmapause, but more consistent with local amplification by substorm‐injected electrons. At higher frequencies the average wave intensities in this sector exhibit two peaks, one near the magnetic equator and one at high latitudes, 45o < |λm| < 60o, with a minimum at intermediate latitudes, 30o < |λm| < 40o, consistent with a source from chorus outside the plasmapause. In the pre‐midnight sector, the intensity of plasmaspheric hiss in the frequency range 50‐1000 Hz decreases with increasing geomagnetic activity. The source of this weak pre‐midnight plasmaspheric hiss is likely to be chorus at larger L∗ in the post‐noon sector that enters that plasmasphere in the post‐noon sector and subsequently propagates eastward in MLT.
30 September 2013 in Journal of Geophysical Research: Space Physics
Global models of the Van Allen radiation belts usually include resonant wave‐particle interactions as a diffusion process, but there is a large uncertainty over the diffusion rates. Here we present a new diffusion matrix for whistler mode chorus waves that can be used in such models. Data from seven satellites are used to construct 3536 power spectra for upper and lower band chorus for 1.5≤L∗≤10 MLT, magnetic latitude 0°≤|λm|≤60° and five levels of Kp. Five density models are also constructed from the data. Gaussian functions are fitted to the spectra and capture typically 90% of the wave power. The frequency maxima of the power spectra vary with L∗ and are typically lower than that used previously. Lower band chorus diffusion increases with geomagnetic activity and is largest between 21:00 and 12:00 MLT. Energy diffusion extends to a few megaelectron volts at large pitch angles >60° and at high energies exceeds pitch angle diffusion at the loss cone. Most electron diffusion occurs close to the geomagnetic equator (<12°). Pitch angle diffusion rates for lower band chorus increase with L∗ and are significant at L∗=8 even for low levels of geomagnetic activity, while upper band chorus is restricted to mainly L∗<6. The combined drift and bounce averaged diffusion rates for upper and lower band chorus extend from a few kiloelectron volts near the loss cone up to several megaelectron volts at large pitch angles indicating loss at low energies and net acceleration at high energies.
24 July 2008 in Space Weather
Département Environnement Spatial, Office National d'Etudes et de Recherches Aérospatiales (ONERA) has been developing a model for the geostationary electron environment since 2003. Until now, this model was called Particle ONERA‐LANL Environment (POLE), and it is valid from 30 keV up to 5.2 MeV. POLE is based on the full complement of Los Alamos National Laboratory geostationary satellites, covers the period 1976–2005, and takes into account the solar cycle variation. Over the period 1976 to present, four different detectors were flown: charged particle analyzer (CPA), synchronous orbit particle analyzer (SOPA), energetic spectra for particles (ESP), and magnetospheric plasma analyzer (MPA). Only the first three were used to develop the POLE model. Here we extend the energy coverage of the model to low energies using MPA measurements. We further include the data from the Japanese geostationary spacecraft, Data Relay Test Satellite (DRTS). These data are now combined into an extended geostationary electron model which we call IGE‐2006.
26 January 2018 in Space Weather