Forecasting the ionospheric f0F2 parameter one hour ahead using a support vector machine technique

This paper proposes a method for forecasting the ionospheric critical frequency, f0F2, 1 h in advance, using the support vector machine (SVM) approach. The inputs to the SVM network are the time of day, seasonal information, 2 month running mean sunspot number (R2), 3 day running mean of the 3 h planetary magnetic ap index, the solar zenith angle, the present value f0F2(t) and its first and second increments, the observation of f0F2 at t−23 h, the 30-day mean value at time, t, fmF2 (t) and the previous 30 day running mean of f0F2 at t−23 h fmF2(t−23). The output is the predicted f0F2 1 h ahead. The network is trained to use the ionospheric sounding data at Haikou, Guangzhou, Chongqing, Lanzhou, Beijing, Changchun and Manzhouli stations at high and low solar activities. The performance of the SVM model was verified with observed data. It is shown that the predicted f0F2 has good agreement with the observed f0F2. The performance of the SVM model is superior to that of the autocorrelation and persistence models, and that it is comparable to that of the neural network model.

Research highlights
► A brief description of the support vector machine is given.
► The SVM method can be used to forecast the ionospheric f0F2.
► The predicted f0F2 has good agreement with the observed f0F2.