Research on Evaporation Duct Prediction Model Based on the GRU Algorithm

The evaporation duct is a kind of atmospheric duct phenomenon, which has an important influence on the efficiency of electromagnetic systems such as radar or radio communication systems. The traditional evaporation duct model depends heavily on the sensitivity and accuracy of the model's parameters. Therefore, we used a data-driven Gated Recurrent Unit (GRU) algorithm, which can predict meteorological elements more easily, quickly and accurately. In this paper, the parameter sensitivity analysis of the classical evaporation duct model, such as P-J model and NPS model, was carried out. Then, the ERA5 reanalysis data provided by the European Center and a single station radiosonde data from 2020 to 2021 were adopted to predict the evaporation duct height by the Gated Recurrent Unit algorithm, which was compared with other machine learning algorithms (such as the Bayes and the SVR). The experimental results show that the NPS model is more stable than the P-J model, and the Gated Recurrent Unit algorithm may have the best prediction accuracy among the other machine learning algorithms.