Cloud top Height Estimation from the Meteosat Water Vapor Imagery Using Computational Intelligence Techniques: Preliminary Results

摘要

This study investigates the cloud top height estimation using nonlinear methods to Meteosat imagery. The suggested approach aims to develop an integrated statistical methodology to estimate the cloud top height on a pixel basis using Meteosat Second Generation water vapor imagery. Radiosonde measurements are used as reference dataset and a spatio-temporal correlation with Meteosat images is performed in order to collect a representative sample for the statistical analysis. Here, we apply Multi Layer Perceptron (MLP) and Support Vector Machines (SVM) and we compare the results to the Linear Regression model. The best results are achieved using SVM for regression. The proposed approach is very promising as it can be used for future in-depth analysis so as to develop a robust approach for geometrical height estimation on a pixel basis of the operational data of Meteosat imagery. It is noted that an accurate estimation of cloud top height can help to eliminate geometric restrictions (e.g. Parallax phenomenon) of the Meteosat satellite imagery, improving its usefulness in a wide area of applications and especially in satellite-based weather forecast.