Abstract
Ionospheric Total Electron Content (TEC) prediction has important
reference significance for the accuracy of global navigation satellite
systems (GNSS) based global positioning system, satellite communications
and other space communications applications. In the study, an available
prediction model of global IGS-TEC map is established based on testing
several different LSTM-based algorithms. We find that Multi-step
auxiliary algorithm based prediction model performs the best. It can
precisely predict the global ionospheric IGS-TEC in the next 6 days (the
MAD and RMSE are 2.485 and 3.511 TECU, respectively). Then, the
autoencoder network algorithm is adopted to construct an assimilation
model that transforming IGS-TEC map to MIT-TEC map. In order to judge
the validity of the assimilation model, the outputs of the assimilation
model are evaluated and compared with the IRI2016 model in four
different geomagnetic storm events. It seems that the assimilation model
can accurately forecast MIT-TEC by inputting the predicted IGS-TEC
value. The performance of assimilation model for the predicting MIT-TEC
is better than that of IRI2016.