Japan Geoscience Union Meeting 2023

Presentation information

[E] Online Poster

P (Space and Planetary Sciences ) » P-EM Solar-Terrestrial Sciences, Space Electromagnetism & Space Environment

[P-EM09] Space Weather and Space Climate

Fri. May 26, 2023 10:45 AM - 12:15 PM Online Poster Zoom Room (2) (Online Poster)

convener:Ryuho Kataoka(National Institute of Polar Research), Antti A Pulkkinen(NASA Goddard Space Flight Center), Mary Aronne, Satoko Nakamura(Institute for Space-Earth Environmental Research, Nagoya University)

On-site poster schedule(2023/5/25 17:15-18:45)

10:45 AM - 12:15 PM

[PEM09-P11] Development of an outer radiation belt forecast model with XAI

*Yuta Nishimiya1, Yoshizumi Miyoshi1, Tomoaki Hori1, Satoko Nakamura1, Masafumi Shoji1, Chae-Woo Jun1, Takefumi Mitani2, Iku Shinohara2, Kazushi Asamura2, Nana Higashio2, Shinji Saito3, Daikou Shiota3 (1.Institute for Space-Earth Environmental Research,Nagoya University, 2.Japan Aerospace Exploration Agency,Institute of Space and Astronautical Science, 3.National Institute of Information and Communications Technology)


The radiation belt is the region in the inner magnetosphere where the most energetic electrons are trapped by the Earth's magnetic field. The spatial and temporal variations of the electron flux are large particularly in the outer radiation belt, and a sustained large flux of the outer belt electrons may lead to satellite anomaly. The prediction of energetic electron flux variation is therefore of significant importance in mitigating these risks. We have developed an outer radiation belt electron forecast model that utilizes Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) techniques. This model is designed to forecast the time-varying electron flux in the outer radiation belt at energies ranging from several hundred keV to several MeV at L = 4-6, based on input data, such as observed solar wind velocity, IMF, and the outer radiation belt electron flux for last three days. The model succeeds in predicting the electron flux in most case; however, it tends to give less accurate flux values when the flux decreases at high L positions. To address this issue, we have improved the model by incorporating additional solar wind parameters. Previous research has indicated that the solar wind dynamic pressure contributes to the loss of energetic electrons via magnetopause shadowing, and thus including this parameter as an input may enhance the forecast accuracy. Our revised model that incorporates the solar wind dynamic pressure as an input demonstrates improved prediction accuracy. Moreover, we have also incorporated Explainable Artificial Intelligence (XAI) into the model to investigate the impact of each input parameter on the electron flux.