Application of Recurrent Neural Network to Modeling Earth's Global Electron Density. Issue 9 (6th September 2022)

• Record Type: Journal Article
• Authors: Huang, Sheng; Li, Wen; Shen, Xiao‐Chen; Ma, Qianli; Chu, Xiangning; Ma, Donglai; Bortnik, Jacob; Capannolo, Luisa; Nishimura, Yukitoshi; Goldstein, Jerry
• Is Part Of: Journal of geophysical research. Volume 127:Issue 9(2022)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2022-09-06
• Subjects: electron density -- deep learning -- LSTM -- encoder‐decoder -- plume -- plasmasphere; Magnetospheric physics -- Periodicals; Space environment -- Periodicals; Cosmic physics -- Periodicals; Planets -- Atmospheres -- Periodicals; Heliosphere (Astrophysics) -- Periodicals; Geophysics -- Periodicals; 523.01
• DOI: 10.1029/2022JA030695
• Languages: English
• ISSNs: 2169-9380

Abstract

Abstract: The total electron density is a fundamental quantity in the Earth's magnetosphere and plays an important role in a number of physical processes, but its dynamic global evolution is not fully quantified yet. We present an implementation of a specific type of recurrent neural network (encoder‐decoder), which is distinct from previous models, to construct global electron density based on the multiyear data from Van Allen Probes. The history of geomagnetic indices is first encoded into a hidden state H, then together with auxiliary information (satellite location), they are decoded into the quantity of interest (total electron density in this study). In this process the input of historical geomagnetic indices is detangled from the satellite location and is processed chronologically by the encoder. As a result, time evolution of geomagnetic indices is explicitly embedded in the structure and the encoded hidden state H can be viewed as the representation of the inner magnetospheric state. The magnetospheric state is then decoded to predict global electron density evolution. Our results show that the model can capture the dynamical evolution of total electron density with the formation and evolution of stable and evident plume configurations that roughly agree with global observations. Our findings demonstrate the importance of applying recurrent neural networks to specify the inner magnetospheric state in a novel way, which will potentially improve our fundamental understanding of wave and particle dynamics in the Earth's magnetosphere. Plain Language Summary: The global evolution of total electron density in the Earth's magnetosphere is important for understanding energetic particle dynamics and their associated wave‐particle interactions. Although physics‐based models show reasonable evolution, they do not fully capture all the dynamical evolution on a global scale. Therefore, data‐driven methods by utilizing machine learning techniques have been developed recently. The recurrent neural network is a variant of artificial neural networks, and it has been applied to many time series data. It naturally handles the time dependence by processing the data chronologically. We develop an encoder‐decoder model based on recurrent neural network and show that by properly designing the architecture, the model can separate the spacecraft orbital variation and the temporal evolution. We apply it to global total electron density modeling trained on the Van Allen Probes observation, and the result shows fairly good performance with stable and evident spatial structures of the plasmasphere and plumes. Key Points: An encoder‐decoder model (recurrent neural network) is developed to reconstruct Earth's global electron density By separating geomagnetic indices from satellite location as inputs, we directly model total electron density with geomagnetic indices The model captures the spatial variation of density by predicting stable and evident plumes, roughly consistent with global observations