Deep Learning Insights Into Ionospheric Sporadic E Irregularities Under Different Solar Activity Conditions

Abstract

Abstract Solar activity profoundly modulates many atmospheric coupling systems, particularly the morphology of the ionospheric irregularities, which is crucial for reliable radio communication. However, the understanding of its long‐term response behavior under different solar activity conditions remains limited, and the exact evolutionary mechanisms of global scale ionospheric coupling system remain poorly constrained. Here we initially show from 2007 to 2018, the weak ionospheric sporadic E scintillation in high latitude regions experienced intensification corresponding with periods of high solar activity. Subsequently, we extend a deep learning model framework called SELF‐ANN proposed by Tian to further clarify the underlying relationship between solar activity and ionospheric irregularities. Compared to SELF‐ANN, we expand the scope of analysis to include data spanning from 2007 to 2018, thus broadening the investigation into the impacts of solar variability. Using long‐term solar activity data and observations of the ionospheric E region irregularity from COSMIC RO, this model was trained to provide solar‐dependent ionospheric morphology. Based on the model, we have successfully achieved high‐precision modeling of weak ionospheric E region irregularities, which commonly occur and are challenging to detect with ionosondes, under different solar activity conditions. Quantitatively, the model achieves a mean absolute error of 0.004, coupled with a Spearman's R value of 0.589 for weak scintillation. Ionospheric reconstructions under different solar activity conditions can improve the understanding of ionospheric evolutionary mechanisms and underscore the importance of incorporating solar variations into ionosphere changes forecasting to ensure the development of future reliable communications.

Keywords

• neural networks
• deep learning
• sporadic E layer
• ionospheric irregularities