Neural-network-based prediction techniques for single station modeling and regional mapping of the <I>fo</I>F2 and M(3000)F2 ionospheric characteristics

Abstract

Abstract. In this work, Neural-Network-based single-station hourly daily foF2 and M(3000)F2 modelling of 15 European ionospheric stations is investigated. The data used are neural networks and hourly daily values from the period 1964-1988 for training the neural networks and from the period 1989-1994 for checking the prediction accuracy. Two types of models are presented for the F2-layer critical frequency prediction and two for the propagation factor M(3000)F2. The first foF2 model employs the E-layer local noon calculated daily critical frequency (foE 12 ) and the local noon F2-layer critical frequency of the previous day. The second foF2 model, which introduces a new regional mapping technique, employs the Juliusruh neural network model and uses the Elayer local noon calculated daily critical frequency (foE 12 ), and the previous day F2-layer critical frequency measured at Juliusruh at noon. The first M(3000)F2 model employs the E-layer local noon calculated daily critical frequency (foE 12 ), its ±3 h deviations and the local noon cosine of the solar zenith angle (cos χ 12 ). The second model, which introduces a new M(3000)F2 mapping technique, employs Juliusruh neural network model and uses the E-layer local noon calculated daily critical frequency (foE 12 ), and the previous day F2-layer critical frequency measured at Juliusruh at noon.

A new global model for the ionospheric F2 peak height for radio wave propagation

“…Xenos (2002) demonstrated the NN technique for single station modelling and regional mapping of M(3000)F2 in the European region.…”

Towards the development of a new global foF2 empirical model using neural networks

“…This has been described by several authors (Williscroft and Poole, 1996;Tulunay et al, 2000;Poole and Poole, 2002). Several other authors have also employed the use of two hidden layers based on the complexity of their networks (Mehrotra et al, 1991;Lamming and Cander, 1999;Derong et al, 2002;Xenos, 2002). On this note, two hidden layers are employed in this work based on the observations that it learned faster and performed better than one hidden layer on the basis of their RMS errors.…”

“…The application of neural networks (NNs) is generally motivated by its principal ability to generalize from a set of training patterns. Several authors have demonstrated that NNs can be used to describe nonlinear phenomena (Wu and Lundstedt, 1996;Williscroft and Poole, 1996;Altinay et al, 1997;Kumluca et al, 1999;Wintoft and Gander, 0273-1177 , 2000Zakharov and Tyrnov, 1999;Mckinnell and Poole, 2000;Xenos, 2002).…”

A single‐station empirical model for TEC over the Antarctic Peninsula using GPS‐TEC data

“…been frequently used in single-station ionospheric empirical models [Francis et al, 2001;Huang and Yuan, 2014;Oyeyemi et al, 2005;Tulunay et al, 2006;Xenos, 2002].…”

“…However, few studies on single‐station TEC empirical models using the nonlinear least squares fitting method have been conducted. Other methods, such as neural networks, have been frequently used in single‐station ionospheric empirical models [ Francis et al , ; Huang and Yuan , ; Oyeyemi et al , ; Tulunay et al , ; Xenos , ].…”