|
Prediction of Ionospheric Scintillation Using Combination of Deep Learning Algorithms in South America
|
Alireza Atabati   , Iraj Jazireeyan * , Mohammad Mahdi Alizadeh Elizeie |
| K. N. Toosi University of Technology |
|
|
Abstract: (441 Views) |
| Ionospheric plasma disturbances, often caused by solar activity and interplanetary magnetic fields, lead to irregular variations in the electron density of the ionosphere, resulting in ionospheric scintillations. The Ionospheric scintillations cause significant fluctuations in the intensity and phase of radio signals, potentially impacting the accuracy of satellite-based navigation systems. This phenomenon occurs irregularly, with higher incidence rates at low latitudes and near the equatorial anomaly compared to mid and high latitudes. This study employs a combination of deep learning network methods and the Huber loss function with observation weighting to predict the spatiotemporal behavior of ionospheric scintillations in near-real-time. Specifically, the ConvGRU method, which combines a GRU network with a convolutional model, is used for spatiotemporal prediction of ionospheric scintillation datasets. The data obtained from 121 ground-based GNSS stations, within the latitude range of 15°N to 55°S and longitude range of 270°E to 330°E across the equatorial anomaly region in South America, are utilized for January, March, and July of the years 2015 (a year of high solar activity) and 2020 (a year of low solar activity). This selection demonstrates the proposed model's accuracy across a diverse spectrum of geomagnetic scenarios. The designed model achieves regional prediction accuracy of approximately 75% for 2015 and 80% for 2020. This study illustrates that advanced machine learning techniques can accurately predict ionospheric scintillations, which can be instrumental in developing near-real-time warning systems for navigation system users, thereby enhancing the precision of observations. |
|
| Keywords: Global Navigation Satellite System (GNSS), Ionospheric Irregularities, Deep Learning Algorithms, Convolutional Gated Recurrent Unit (ConvGRU), Huber Loss Function, Amplitude Ionospheric Scintillation (S4) |
|
|
|
Type of Study: Research |
Subject:
Geodesy
Received: 2024/04/23 | Accepted: 2024/08/3 | ePublished ahead of print: 2025/03/17
|
|
|
|
|
|
|
| Send email to the article author |
|
|
