PCR Model: Optimization of Evaporation Duct Height Prediction Based on Principal Component Regression

Evaporation duct is an abnormal refraction atmospheric structure frequently appearing on the sea surface. Accurately obtaining the evaporation duct height (EDH) has significant effects on the effective application of electromagnetic system equipment. Aiming at the problems of large error and strong data dependence of current evaporation duct prediction models, we take the measured meteorological observation data as the starting point and propose the principal component regression evaporation duct prediction Model (PCR model) which applies the principal component regression (PCR) in this field for the first time. We select the famous theoretical Paulus-Jeske (PJ) model and machine learning Gradient Boosting Regression (GBR) model as the baseline methods, and conduct the prediction accuracy comparison experiment, cross-prediction experiment, and PCR model optimization experiment successively. The results show that compared with the GBR model, the EDH prediction accuracy of the PCR model is significantly improved in five sub-areas and the generalization ability is also obviously enhanced. The accuracy of the PCR model optimized by combining reanalysis data is greatly improved compared with its unoptimized version.