Multi-Model Ensemble Diagnostic Method for Evaporation Duct Considering Physical Property

Evaporation duct is an phenomenon in the marine atmosphere that can cause anomalous transmission of electromagnetic (EM) waves. Therefore, accurately obtaining the evaporation duct height (EDH) is of great significance to the propagation of EM waves. Currently, evaporation duct diagnostic methods based on machine learning only consider mathematical relationships of data, without considering the physical properties of ocean-atmosphere interaction. To overcome the above problems, a multi-model ensemble diagnostic method for evaporation duct considering physical properties with convolutional neural network (CNNM) is proposed in the paper. Using meteorological and hydrological parameters and optimal model diagnostic values as inputs and EDH reference values as outputs constructing a deep learning model-CNNM. Experiments and analyses of the diagnostic method were conducted using the observational data from 19 August 2023 to 16 September 2023 on the research vessel “Tan Kah Kee”. The results show that the CNNM is more suitable for EDH diagnosis and has a higher accuracy than other models. Under stable atmospheric condition, the average errors of diagnostic values of CNNM model from the reference values is 1.09 m. Under unstable atmospheric condition, the average errors of diagnostic values of CNNM model from the reference values is 1.06 m.