PINE-RT: An operational real-time plasmasphere model

Abstract

The plasmasphere is a region of cold and dense plasma around the Earth, corotating with the Earth. Its plasma density is very dynamic under the influence of the solar wind and it influences several processes such as the GPS navigation, the surface charging of the satellites and the propagation and growth of plasma waves. In this manuscript, we present a new machine-learning model of the equatorial plasma density depending only on the Kp index and the solar-wind properties at the L1 Lagrange point. We call this model PINE-RT as it has been inspired by the recently-introduced PINE (Plasma density in the Inner magnetosphere Neural network-based Empirical) model and it has been developed to run in real-time (RT) in the context of the PAGER project. This project is an EU Horizon 2020 project aiming at forecasting the threats of satellite charging as a consequence of the solar activity 1–2 days ahead. In PAGER, the Kp index and the solar-wind properties at L1 are the inputs which are made available for the plasmasphere modeling. We report here the detailed derivation of the PINE-RT model and its validation and comparison with two state-of-the-art machine-learning and physics-based models. The model is currently running in real-time and its predictions are publicly available.

Global validation of data-assimilative electron ring current nowcast for space weather applications

“…The four-dimensional Versatile Electron Radiation Belt code (VERB-4D) 48 describes the evolution of phase space density with a convective-diffusive equation in MLT, radial distance R , and the two modified adiabatic invariants V and K 49 : where and J represent the first and second adiabatic invariants 50 , and is the rest mass of an electron. VERB-4D has been extensively used for simulating radiation belts 48 , 51 , plasmasphere 38 , 52 , and ring current dynamics 53 .…”

Electron Density Specification in the Inner Magnetosphere From the Narrow Band Receiver Onboard DSX

“…Recently, the fast expansion of neural networks has enabled the development of machine learning-based plasmaspheric models (Ace, 2021;Bianco et al, 2023;Chu et al, 2017aChu et al, , 2017bHuang et al, 2022;Zhelavskaya et al, 2017Zhelavskaya et al, , 2018Zhelavskaya et al, , 2021 to predict global plasma density in the inner magnetosphere. The accuracy of these models was found to be greatly dependent on the training data quality and availability.…”