Bidirectional LSTM for ionospheric vertical Total Electron Content (TEC) forecasting

The ionosphere is a region over earth's atmosphere which is ionized by solar radiation. It plays an important part in atmospheric electricity and forms the inner edge of the magnetosphere, furthermore, it has practical importance for its effects on radio propagation to earth. Accordingly, its cyclic changes and disturbances could influence communication, navigation, radar seriously. Total Electron Content (TEC) is an important parameter reflecting ionospheric significant characteristic. We can analyse ionospheric disturbance and cyclic change by investigating TEC. Now, the time sequence of TEC and relevant parameters is available for our research. In addition, Bidirectional long short-term memory (Bi-LSTM) is believed to have potential in time sequence processing. Hence, we investigate TEC forecast by referring to Bi-LSTM in this work. Experimental results demonstrate that Bi-LSTM can capture the cyclic change feature of TEC, and perform better than LSTM and multi-LSTM for TEC forecast.