72‐Hour Time Series Forecasting of Hourly Relativistic Electron Fluxes at Geostationary Orbit by Deep Learning

Abstract

In this study, we forecast hourly relativistic (>2 MeV) electron fluxes at geostationary orbit for the next 72 hr using a deep learning model based on multilayer perceptron. The input data of the model are solar wind parameters (temperature, density and speed), interplanetary magnetic field (|B| and Bz), geomagnetic indices (Kp and Dst), and electron fluxes themselves. All input data are hourly averaged ones for the preceding 72 consecutive hours. We use electron flux data from Geostationary Operational Environmental Satellite (GOES)‐15 and ‐16, and perform a mapping for matching these two data. Total period of the data is from 2011 January to 2021 March (GOES‐15 data for 2011–2017 and GOES‐16 data for 2018–2021). We divide the data into training set (January–August), validation set (September), and test set (October–December) to consider the solar cycle effect. Our main results are as follows. First, our model successfully predicts hourly electron fluxes for the next 72 hr. Second, root‐mean‐square error of our model is from 0.18 (for 1 hr prediction) to 0.68 (for 72 hr prediction), and prediction efficiency is from 0.97 to 0.53, which are much better than those of the previous studies. Third, our model well predicts both diurnal variation and sudden increases of electron fluxes associated with fast solar winds and interplanetary magnetic fields. Our study implies that the deep learning model can be applied to forecasting long‐term sequential space weather events.

A data assimilation-based forecast model of outer radiation belt electron fluxes

“…The figure shows that as the prediction time increases, the model performance gradually declines, regardless of electron energy and L*. Son et al (2022) developed a deep learning model to predict the fluxes of >2 MeV electrons at GEO and PE for 1, 2, and 3 days ahead predictions were 0.78, 0.64, and 0.53, respectively. Although PE at GEO in our model is not available because of the absence of satellite measurements, Figure 5 shows that our model performance for ~0.3-3 MeV electrons at L* = ~3.25-4.5 is comparable or even better than that of Son et al (2022).…”

“…Boynton et al (2016) developed an empirical model based on the Nonlinear Autoregressive Moving Average with Exogenous inputs (NARMAX) algorithm to predict geosynchronous electron fluxes. Recently, the neural network technique has been extensively utilized in predicting electron fluxes in the radiation belts (Ling AG et al, 2010;Shin et al, 2016;Wei LH et al, 2018;Zhang H et al, 2020;Son et al, 2022;Tang RX et al, 2022;Wing et al, 2022;Wang JH et al, 2023).…”

A Modeling Study of ≥2 MeV Electron Fluxes in GEO at Different Prediction Time Scales Based on LSTM and Transformer Networks

“…With the rapid progress of artificial intelligence, machine learning models have been applied to predict high-energy electron fluxes at GEO orbit. Many machine-learning methods have also been ⋆ Just to show the usage of the elements in the author field applied, such as Fukata et al (2002), Ukhorskiy et al (2004), Xue and Ye (2004), Ling et al (2010), Balikhin et al (2011), Balikhin et al (2016), Wei et al (2011), Wang and Shi (2012), Boynton et al (2013), Boynton et al (2015), Guo et al (2013), Pakhotin et al (2014), Ganushkina et al (2014), Ganushkina et al (2015), Shin et al (2016), Wei et al (2018), Zhang et al (2020), Katsavrias et al (2022), Landis et al (2022), Saikin et al (2021), and Son et al (2022). Wei et al (2018) Observation data from satellites often contain data gaps for ≥2 MeV electron fluxes.…”

Prediction of Radiation Belts Electron Fluxes at a Low Earth Orbit Using Neural Networks With PROBA‐V/EPT Data

“…Many works have arisen these last two decades where the satellite measurements in the inner magnetosphere have been used to figure out radiation belt dynamics and, more recently, to train machine learning models. The Geostationary Operational Environmental Satellite network with its fleet of geosynchronous equatorial orbit (GEO) satellites launched since the seventies (https://www.nasa.gov/content/goes) have been extensively employed for such goal (see e.g., Landis et al., 2022; Myagkova et al., 2019; Son et al., 2022; Sun et al., 2021; Wei et al., 2018). The launch of Radiation Belt Storm Probes (RBSP) in 2012 in a Low‐Medium Earth orbit with a highly elliptic equatorial trajectory (Mauk et al., 2012), later renamed Van Allen Probes, has enormously contributed to the research on the physical processes governing the radiation belts and revamped our knowledge of the near‐Earth space by demonstrating, for example, the existence of a third ultra‐relativistic radiation belt (Mann et al., 2016) during high geomagnetic activity.…”