LARRES: A New Deep Learning Based Global Ionosphere Map Prediction Model with Large Receptive Field

The LARRES model was proposed to address the issue of global ionosphere prediction. Based on its large receptive field structure, it can take into account both local features and the global features of the Global Ionosphere Map (GIM), thereby significantly enhancing the accuracy of GIM prediction. In this paper, the performance of LARRES was validated using eight-year GIM data. The experimental results indicate that LARRES is well-suited to the characteristics of global ionospheric variations. Compared with previous models, LARRES achieves a remarkable improvement in accuracy, not only in terms of overall precision but also across different months and latitudes. LARRES can effectively overcome the limitations of the ConvLSTM model in handling global features. Given that current research on GIM prediction is largely confined to modifications of the ConvLSTM model, our study breaks through these constraints and paves the way for a new research direction.