Local TEC modelling and forecasting using neural networks

Abstract Modelling the Earth's ionospheric characteristics is the focal task for the ionospheric community to mitigate its effect on the radio communication, and satellite navigation. However, several aspects of modelling are still challenging, for example, the storm time characteristics. This paper presents modelling efforts of TEC taking into account solar and geomagnetic activity, time of the day and day of the year using neural networks (NNs) modelling technique. The NNs have been designed with GPS-TEC measured data from low and mid-latitude GPS stations. The training was conducted using the data obtained for the period from 2011 to 2014. The model prediction accuracy was evaluated using data of year 2015. The model results show that diurnal and seasonal trend of the GPS-TEC is well reproduced by the model for the two stations. The seasonal characteristics of GPS-TEC is compared with NN and NeQuick 2 models prediction when the latter one is driven by the monthly average value of solar flux. It is found that NN model performs better than the corresponding NeQuick 2 model for low latitude region. For the mid-latitude both NN and NeQuick 2 models reproduce the average characteristics of TEC variability quite successfully. An attempt of one day ahead forecast of TEC at the two locations has been made by introducing as drivers previous day solar flux and geomagnetic index values. The results show that a reasonable day ahead forecast of local TEC can be achieved.

[1]  John Bosco Habarulema,et al.  Regional GPS TEC modeling; Attempted spatial and temporal extrapolation of TEC using neural networks , 2011 .

[2]  Sandro M. Radicella,et al.  A new version of the NeQuick ionosphere electron density model , 2008 .

[3]  Jan Laštovička,et al.  Solar activity impact on the Earth’s upper atmosphere , 2013 .

[4]  John M. Goodman,et al.  Operational communication systems and relationships to the ionosphere and space weather , 2005 .

[5]  Sandro M. Radicella,et al.  TEC ingestion into NeQuick 2 to model the East African equatorial ionosphere , 2012 .

[6]  Peter Wintoft,et al.  Ionospheric foF2 storm forecasting using neural networks , 2000 .

[7]  Lyubka Pashova,et al.  Global TEC maps based on GNSS data: 1. Empirical background TEC model , 2013 .

[8]  Hong Yuan,et al.  Ionospheric single‐station TEC short‐term forecast using RBF neural network , 2014 .

[9]  David J. C. MacKay,et al.  A Practical Bayesian Framework for Backpropagation Networks , 1992, Neural Computation.

[10]  Keith M. Groves,et al.  Validation of the NeQuick 2 and IRI-2007 models in East-African equatorial region , 2013 .

[11]  Sandro M. Radicella,et al.  A near‐real‐time model‐assisted ionosphere electron density retrieval method , 2006 .

[12]  S. Basu,et al.  Effects of Large Magnetic Storms on Communication and GPS Navigation Systems at Middle and Equatorial Latitudes , 2001 .

[13]  Ljiljana R. Cander,et al.  Artificial neural network applications in ionospheric studies , 1998 .

[14]  Lee-Anne McKinnell,et al.  Application of neural networks to South African GPS TEC modelling , 2009 .

[15]  Lee-Anne Williscroft,et al.  Neural networks, foF2, sunspot number and magnetic activity , 1996 .

[16]  Iwona Stanislawska,et al.  An Overview of Ionosphere—Thermosphere Models Available for Space Weather Purposes , 2009 .

[17]  Simon Haykin,et al.  Neural Networks: A Comprehensive Foundation , 1998 .

[18]  Michael J. A. Berry,et al.  Data mining techniques - for marketing, sales, and customer support , 1997, Wiley computer publishing.

[19]  B. Hofmann-Wellenhof,et al.  Global Positioning System , 1992 .

[20]  Sandro M. Radicella,et al.  Data ingestion into NeQuick 2 , 2011 .

[21]  Adam Blum,et al.  Neural Networks in C++: An Object-Oriented Framework for Building Connectionist Systems , 1992 .

[22]  R.K. Jurgen,et al.  Sarnoff Labs: 'still crazy' but coping , 1988, IEEE Spectrum.

[23]  Lee-Anne McKinnell,et al.  Prediction of global positioning system total electron content using Neural Networks over South Africa , 2007 .

[24]  Michael Y. Hu,et al.  Forecasting with artificial neural networks: The state of the art , 1997 .

[25]  Lee-Anne McKinnell,et al.  TEC measurements and modelling over Southern Africa during magnetic storms; a comparative analysis , 2010 .

[26]  Sandro M. Radicella,et al.  Calibration errors on experimental slant total electron content (TEC) determined with GPS , 2007 .

[27]  R. Hecht-Nielsen,et al.  Neurocomputing: picking the human brain , 1988, IEEE Spectrum.

[28]  Kurt Hornik,et al.  Multilayer feedforward networks are universal approximators , 1989, Neural Networks.

[29]  Mehdi Khashei,et al.  An artificial neural network (p, d, q) model for timeseries forecasting , 2010, Expert Syst. Appl..

[30]  Kristen Bell DeTienne,et al.  Neural Networks as Statistical Tools for Business Researchers , 2003 .

[31]  Norbert Jakowski,et al.  Comparative testing of four ionospheric models driven with GPS measurements , 2011 .

[32]  Sandro M. Radicella,et al.  The NeQuick model genesis, uses and evolution , 2009 .

[33]  E. T. Senalp,et al.  Forecasting total electron content maps by neural network technique , 2006 .

[34]  Lee-Anne McKinnell,et al.  On the predictability of f0F2 using neural networks , 2000 .

[35]  Takashi Maruyama,et al.  Regional reference total electron content model over Japan based on neural network mapping techniques , 2007 .

[36]  Dave Winkler,et al.  Bayesian Regularization of Neural Networks , 2009, Artificial Neural Networks.