Forecasting of Global Ionosphere Maps With Multi‐Day Lead Time Using Transformer‐Based Neural Networks

Abstract

Ionospheric total electron content (TEC) is a key indicator of the space environment. Geophysical forcing from above and below drives its spatial and temporal variations. A full understanding of physical and chemical principles, available and well‐representable driving inputs, and capable computational power are required for physical models to reproduce simulations that agree with observations, which may be challenging at times. Recently, data‐driven approaches, such as deep learning, have therefore surged as means for TEC prediction. Owing to the fact that the geophysical world possesses a sequential nature in time and space, Transformer architectures are proposed and evaluated for sequence‐to‐sequence TEC predictions in this study. We discuss the impacts of time lengths of choice during the training process and analyze what the neural network has learned regarding the data sets. Our results suggest that 12‐layer, 128‐hidden‐unit Transformer architectures sufficiently provide multi‐step global TEC predictions for 48 hr with an overall root‐mean‐square error (RMSE) of ∼1.8 TECU. The hourly variation of RMSE increases from 0.6 TECU to about 2.0 TECU during the prediction time frame.

Deep Learning‐Based Prediction of Global Ionospheric TEC During Storm Periods: Mixed CNN‐BiLSTM Method

“…Moreover, some studies have utilized Convolutional Long Short-Term Memory neural networks, which incorporate spatial information into LSTM, to predict ionospheric parameters (Liu et al, 2022;Luo et al, 2023;Xia et al, 2022). Some studies used the Transformer to predict TEC, and by increasing the depth of the model, the prediction accuracy of the model was further improved (Shih et al, 2024;Yuan et al, 2023). In addition, nowcasting technique has been improved by Chen et al (2019Chen et al ( , 2021 In conclusion, the advancements of deep learning models such as RNN and CNN have made significant contributions to the prediction of ionospheric parameters.…”

MaxEnt SeismoSense Model: Ionospheric Earthquake Anomaly Detection Based on the Maximum Entropy Principle

“…The advent of the transformer neural network architecture, introduced by Google in 2017, has sparked considerable interest across diverse domains. Post-2022, research on ionospheric prediction leveraging improved transformer-based models has surged, yielding notable successes [45][46][47][48]. While LSTM and GRU, along with their respective optimizations, have demonstrated commendable performance and mitigated long-range dependencies in sequence tasks to a certain extent, their efficacy diminishes notably with increasing sequence lengths.…”