Clear-air anomaly detection using modified Kalman temporal filter from geostationary multispectral data
A multispectral temporal-based remote sensing technique based on a modified Kalman filter is presented for clear-air detection by using Geostationary visible-infrared radiometric passive measurements. The Kalman estimate relies on a model of the daily measurement cycle of the considered pixel in clear-sky conditions. If the measurement significantly deviates from the predicted value, an anomaly is detected, which is interpreted as a non-clear air scenario. The add-on value of such approach is to be able to provide a-priori estimates, making the algorithm applicable in a global way. The Meteosat Second Generation satellite has been used over a large sample area in West Africa and a test period of three months. An inter-comparison with respect to the EUMETSAT cloud mask product has been carried out showing promising results in terms of detecting clear-air scenarios and percentages of matching around 90% over the entire period.