Automatic identification of Spread F using decision trees

Spread F is a commonly observed phenomenon on ionograms caused by plasma irregularities or wave-like structures in the ionosphere. In general, Spread F could induce fluctuations of amplitude and phase of radio waves which travel through the ionosphere. Therefore, investigation of Spread F could be used to not only reveal ionospheric electrodynamics process, but also have a significant engineering application. Due to a large amount of ionograms recorded by ionosondes, it is a challenge work to manually identify ionograms with Spread F to study characteristics of Spread F. Thus, much work has been devoted to automatic identification of Spread F. In the present study, a machine learning method related to decision tree was adopted to automatically identify Spread F from ionograms. First, ionograms were processed by image method and projection techniques to provide input parameters for decision tree. The output of the proposed decision tree is whether Spread F is present or not on ionograms. Then, a set of ionograms was used to construct a decision tree. At last, a set of ionograms was adopted to validate the performance of this decision tree. In this study, ionograms recorded at Puer station (Geographic latitude and longitude: 22.7°N, 101.5°E; Geomagnetic latitude: 12.8°N) in the Yunnan province were used. Results indicate that the decision tree performed well in automatic identification of Spread F on ionograms. The accuracy of automatic identification in a set of ionograms with Spread F was reached up to 89%. It inspires us to continually improve the performance of automatic identification of Spread F in the future work.