Deep learning for ionogram parameters scaling at polar region ionosphere

Ionosondes are most widely used instruments to obtain electron density profiles of ionosophere. Typically ionosondes operates with $\gt 15$ minutes time intervals that is enough to obtain regular parameters of the ionosphere, but insufficient to study small- and medium-scale traveling ionospheric disturbances and sporadic E layers. The key points for such studies are the increase of the ionosondes time resolution, as well as automation of ionogram scaling routine. In this study we show the results of implementation of deep learning for ionogram parameters scaling. We trained and tested a convolutional neural network on data of Sodankyla ionosonde at high latitude region. Our results show a close to human ability to recognise layers shape of F1, F2, E, 6 types of Es and scaling foF1, foF2, foE, foEs parameters.