Atmospheric Numerical Prediction Based on Graph Convolution, Waveguide Discrimination, and Electromagnetic Transmission

Atmospheric-related values are increasingly available from ground-based observatories, providing opportunities for data mining and prediction. This study aims to use artificial intelligence methods to analyze and predict atmospheric values under complex meteorological conditions, focusing on the propagation characteristics of electromagnetic waves, especially within meteorological ducts. The first step of the study was to collect and process data. We obtained simulated output data from the WRF model and actual measured data from 51 IGRA stations, which included key meteorological parameters such as temperature, humidity, and pressure. To understand the meteorological correlations among these stations, we applied the K-means clustering algorithm to classify the data from these 51 stations, identifying regions with similar meteorological characteristics. Subsequently, we employed Graph Convolutional Networks (GCNs) for an in-depth analysis of these time-series data. GCNs effectively managed the complex spatial relationships between stations and predicted future changes in temperature, humidity, and pressure. This was achieved by integrating various neural network models to optimize prediction accuracy, and the forecasted results were validated against actual measured and IGRA observational data for subsequent analysis and study of electromagnetic transmission. After obtaining the predicted data for future meteorological parameters, we further calculated the conditions for the existence of ducts and used ray tracing to simulate the propagation paths of electromagnetic waves under these specific duct conditions. Our model not only considered propagation effects such as direct transmission, reflection, and refraction but also simulated the characteristics of electromagnetic wave propagation in ducts, providing a new perspective on the behavior of electromagnetic waves under different meteorological conditions. This study offers new insights into the integration of artificial intelligence with traditional electromagnetic observation data. Based on this, we combined analysis and prediction to understand the impact of ducts on electromagnetic wave transmission.