Near real time detection of hot spots on Meteosat Second Generation images : from forest fires to volcanic eruptions

In the 2005 IGARSS’ symposium, we presented the setting up of a data receiving station for METEOSAT SECOND GENERATION (MSG) satellites by undergraduates and we emphasized the pedagogical interest of the project. The experience has been successfully pursued by developing with the undergraduates involved the exploitation of the data received and its promotion. Among the different activities implemented in this context, we discuss in this article the work concerning the hot spots detection in near real time on MSG images. This approach, beyond the technical and pedagogical aspects, allows the future engineers to be involved in very important issues such as global warming and natural disasters. In order to define the forest fire detection algorithms adapted to MSG data, we start by describing the physical properties that lead us to identify the more efficient spectral bands available in MSG. Then, we deal with the different algorithms developed and the accuracy and efficiency of the detection compared with other published results. Sometimes, we observed very localised and motionless fires during several days or even several weeks. Looking at more precise scales, we realized that these anomalies are mostly localised on active volcanoes. In order to validate the hypothesis that we really detect volcanic eruptions, we focused our attention on a very specific volcano called Karthala, Grand Comore Island, Indian Ocean. Comparing our results with ground data, we observe a very good detection of the eruption’s activity from the start to the end with almost a quarter hour accuracy.