![]() | Only 14 pages are availabe for public view |
Abstract The Ionosphere, being part of the upper atmosphere of Earth, is a highly variable environment. The ionospheric electron density varies spatially with altitude and latitude, and temporally with a time of day, season and solar activity. The ionosphere affected by the space weather conditions. Space weather presents a threat to radio communications used in military and civilian human activities through impacts on technology in applications such as Aviation, Navigation, Satellite operation and Telecommunication. Radio communications can disturbed by the space weather related changes in the ionospheric parameters specially the totel electron content (TEC). The totel electron content (TEC) is an important ionospheric feature that can be used for several purposes, such as the studies of the ionosphere-plasmasphere system and space weather applications, as well as for Global Navigation Satellite Systems (GNSS) applications. The changes in the ionospheric VTEC affects the accuracy of the on the Global Navigation Satellite System (GNSS) used for Navigation, Positioning and Timing. Therefore, the physics of the ionosphere has to be well studied and, the ionospheric Nowcasting and modeling of the ionosphere could help mitigate some of these disturbances which could result in loss or damage. In addition They are very important for the first step in the topside ionospheric imaging and the key factor of the space weather forecasting. I focused on African sector because it presents a special challenge since there is a lack of topside ionospheric information. This study presents the first prediction results of the neural network (NN) model for the topside ionospheric vertical total electron content obtained from Swarm-A data during the time period from 1 January 2014 to 30 September 2019 over the European-African sector. Five years of data from 2014 – 2018 are combined from Swarm-A and OMNI database of solar and geomagnetic indices to train the neural network (NN) model, and the data of 2019 is used to test the NN performance in predicting the topside ionospheric vertical total electron content. We categorized the data into two main categories; the quiet and disturbed days. After that divided each category into two sub-categories according to the Swarm-A trajectory whether it is ascending or descending, then implemented four NN models: 1) quiet day-ascending (QD Asc), 2) quiet day descending (QD Des), 3) disturbed day ascending (DD Asc), and 4) disturbed day descending (DD Des). I validated the NN models by comparison of the predictions with Swarm-A measurements of VTEC in 2019. Then, compared the performance of the NN models with the IRI2016 model using Swarm measurements as a reference. Finally demonstrated that the NN models give better results than the IRI2016 model by using the root mean square error (RMSE). The predictions of the IRI2016 model frequently underestimates the Swarm-A VTEC. As well as the one hidden layer is better in these cases: QD Asc and QD Des models. In addition, the two double hidden layers perform better in the following cases: DD Asc and DD Des models. Abstract The Ionosphere is a highly variable environment and effects on the space technologies particularly communication systems, therefore, the Nowcasting and modelling of the ionosphere is very important. The total electron content TEC is an important ionospheric feature, this study presents the first prediction results of the neural network model for the topside ionospheric vertical total electron content obtained from Swarm-A data during the time period from 1 January 2014 to 30 September 2019 over the European-African sector. Five years of data from 2014 – 2018 are used to train the NN model, and the data of 2019 is used to test the NN performance in predicting the topside ionospheric vertical total electron content. We implemented four NN models: 1) quiet day-ascending, 2) quiet day-descending, 3) disturbed day-ascending, and 4) disturbed day-descending. We compared Swarm_NN model performance in prediction with IRI2016 model by using the RMSE. Swarm_NN model has showed a good result more than IRI2016 model. Key Words: Swarm Constellation Satellites – Neural Network – IRI2016model –Ionospheric Modelling – Ionospheric Nowcasting |