Deep Learning for Ionogram Parameter Extraction: A Time-Series Approach to Ionospheric Monitoring

In recent decades, various tools have been developed for the extraction of ionospheric parameters with ionograms, for example: ARTIST-5 and among others. There are some aproximations in previous works using deep learning for automatic scaling with parameters extraction of great importance as the identification of the E and F2 layer, as well as the occurrence of the maximum frequency in the F2 layer (foF2) and the height at which it occurs (hmF2). These tools generally work accurately for relatively quiet days (QD), but not for spread-F days. Where a lot of variability is observed in the parameters obtained on those days, where traditional manual correction is necessary. In this work, we trained a model combining CNN, LSTM and Dense layers that is able to capture the short-term variability of the ionosphere and our trained model returns the frequency profile values on 0 to 1000 kilometers for ionograms from VIPIR ionosondes of LISIN Network. We tested and compared our frequency profiles from model with manual scaling and parameters obtained with ARTIST 5.0.