A review of machine learning-based ionospheric spatial and temporal modeling
The ionosphere introduces disturbances and errors in radio signals for satellite communication and navigation, requiring precise modeling to mitigate these effects. However, accurate ionospheric modeling faces challenges due to the complex spatio-temporal variations caused by intricate coupling with...
| Autores: | , , |
|---|---|
| Formato: | artículo |
| Fecha de publicación: | 2025 |
| País: | España |
| Recursos: | Universitat Politècnica de Catalunya (UPC) |
| Repositorio: | UPCommons. Portal del coneixement obert de la UPC |
| Idioma: | inglés |
| OAI Identifier: | oai:upcommons.upc.edu:2117/457127 |
| Acesso em linha: | https://hdl.handle.net/2117/457127 https://dx.doi.org/10.1029/2024JH000555 |
| Access Level: | acceso abierto |
| Palavra-chave: | Machine learning Ionosphere Aprenentatge automàtic Ionosfera Classificació AMS::85 Astronomy and astrophysics Àrees temàtiques de la UPC::Matemàtiques i estadística::Matemàtica aplicada a les ciències |
| Resumo: | The ionosphere introduces disturbances and errors in radio signals for satellite communication and navigation, requiring precise modeling to mitigate these effects. However, accurate ionospheric modeling faces challenges due to the complex spatio-temporal variations caused by intricate coupling with the lower atmosphere and Earth's magnetic field, as well as the influence of space weather events. As a powerful tool capable of uncovering nonlinear relationships between inputs and outputs, machine learning (ML) approaches have been increasingly applied to ionospheric modeling, showing great potential. In this review, we explore studies from the past decades on the application of various ML methods in both spatial and temporal ionospheric modeling. The ionospheric parameters involved include vertical total electron content (VTEC), F2-layer critical frequency (foF2), and virtual height (hmF2). By synthesizing findings from various studies, we present a comprehensive overview of the accuracy achieved by different ML-based ionospheric models. Additionally, we summarize commonly employed data sources, ML algorithms, and strategies for best practices, offering insights and guidance for future research. Beyond highlighting achievements, the review also points out limitations in current studies and provides perspectives for further improving ML models. |
|---|