Machine Learning methods of FY-3C tropopause product correction

This article tried to use machine learning methods to improve the accuracy of Fengyun-3C Meteorological Satellite(FY-3C) Global Navigation Satellite System Occultation Sounder (GNOS)tropopause products to the corresponding value of European Centre for Medium-range Weather Forecasts(ECMWF). The principles of the three methods are introduced, the correction results of the three methods are given. The correction effect of tropopause parameters in high latitude area of the Back Propagation Neural Networks and the Support Vector Regression method is similar. The latitude average of the tropopause parameters after the correction of Random Forest Regression is the best fit with the ECMWF results, and the reduction of the root mean square error is particularly obvious.