Data Sheet 3_CEEMDAN-TVF-EMD-TCN based method for elevated duct parameters predicting with high-resolution radiosonde data.csv

IntroductionAtmospheric ducts represent the most frequently observed refractive phenomenon in maritime atmospheric environments. Prior research has shown that the occurrence of elevated ducts exceeds 60% with high-resolution radiosonde data, providing a solid foundation for developing communication applications based on this propagation mechanism.
MethodsTo support the design of elevated ducts communication systems, this paper proposes a long-term elevated duct thickness prediction model, which integrates a dual decomposition framework consisting of adaptive noise complete ensemble empirical mode decomposition (CEEMDAN) and time-varying filter empirical mode decomposition (TVF-EMD) with a temporal convolutional network (TCN) method. The CEEMDAN-TVF-EMD-TCN elevated duct thickness prediction model (CTETEDP) uses the 2016–2022 datasets at the SAN JUAN station.
ResultsThe CTETEDP model realizes a high-precision time series prediction of duct thickness, with a root mean square error (RMSE) of 16.3 m and a mean absolute error (MAE) of 12.8 m. To further evaluate the adaptability of the CTETEDP method in different regions, predictions were carried out based on the observation data at the MAJURO and TRUK INTL stations, respectively. In the spring, summer, and autumn of 2022, the RMSEs at MAJURO station were 14.3 m, 8.8 m, and 12.4 m, respectively, while those at TRUK INTL station were 29.8 m, 12.3 m, and 17.2 m.
DiscussionThese results indicate that the CTETEDP model has good adaptability and practicality and can realize high-precision and long-term continuous prediction of elevated duct thickness, which will provide support for future elevated duct-based communication systems.