Ionospheric foF2 storm forecasting using neural networks

The ionosphere shows a large degree of variability on time scales from hours to the solar cycle length. This variation is associated with magnetospheric storms, the Earth's rotation, the season, and the level of solar activity. To make accurate predictions of key ionospheric parameters all these variations must be considered. Neural networks, which are data driven non-linear models, are very useful for such tasks. In this work we examine if the F2 layer plasma frequency, foF2, at a single ionospheric station can be predicted 1 to 24 hours in advance by using information of past foF2 observations, magnetospheric activity, and time as inputs to neural networks. Particular attention has been paid to periods when great geomagnetic storms were in progress with the aim to develop a successful ionospheric storm forecasting tool.