Improvement of cloud detection near sunrise and sunset by temporal-differencing and region-growing techniques with real-time SEVIRI

Abstract

The European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT), through the Satellite Application Facility for Nowcasting (SAFNWC), provides a software package suitable to generate real-time products derived from the Spinning Enhanced Visible and Infrared Imager (SEVIRI) on board the Meteosat Second Generation (MSG) satellite. A cloud mask is one of its key derived products. An analysis of the behaviour of version v2008 of this software using animation loops indicates a discontinuity in the detection of low clouds near the day–night transition. Thus, its use may have a negative impact when the scene includes such conditions. Our study provides solutions for a smoother continuity. It shows that a temporal-differencing technique combined with a region-growing technique can improve low-level cloudiness detection in the day–night transition area at the expense of a small increase of false alarms. The method and the resulting improvement are described and illustrated. A comparison of statistics of the cloudiness over Europe reported in Surface Synoptic Observations (SYNOP) and retrieved from the SEVIRI with and without the enhancement estimates a decrease of 50% of the frequency of missed clouds while the false alarm ratio increases only marginally. A comparison with another EUMETSAT operational SEVIRI-based cloud mask confirms the usefulness of the proposed technique at oblique sun angles.

Acknowledgements

This work was carried out under the EUMETSAT SAFNWC programme. We thank our colleagues of the Centre de Météorologie Spatiale for their help in data collection and for making satellite imagery and derived products available in real time for forecasters. Thanks are also due to Chris Merchant (School of GeoSciences, University of Edinburgh, UK) for his correction of the English before the review process. We are also grateful to the IPL98 developers for making their library available under the LGPL licence at http://www.mip.sdu.dk/ipl98.