A reef called ''Les Hermelles'' occurs in the intertidal zone of the Bay of Mont Saint-Michel (Normandy, France). The reef is composed of the sea-worm Sabellaria alveolata and thus, is one of the northernmost biological reefs in the world. To map Les Hermelles in the field is difficult because of its great distance from the coastline and its brief exposure to the air. Consequently, the last map dates back to 1980. In 1984, mussel farms were developed around the reef. Since then, they are often regarded as the factor responsible for the perceived decline of the reef. The main goal of the study was to assess the possibility of mapping Les Hermelles using satellite imagery and if so, to give a rough idea about its recent evolution since the 1980 map. A multi-spectral SPOT image satellite was used for the tests. Areas of reef, sand, silt, mud and seawater show similar spectral signatures. Moreover, these environments are often mixed within a single pixel. Consequently visual interpretation and the usual classification approaches failed, showing inter-class confusion. A new classification based on cellular automata appears to be promising. Applied to the example of Les Hermelles, this method was successful and showed a certain stability of the reef dynamic tendencies since 1980. Although the preliminary results have to be securely based on field data acquired close to the satellite-image date, they already confirm the key role that sandbars play in the reef dynamics whereas the role of the mussel farms remains unclear if not null. The classification by cellular automata presents the advantage that no supervision is required, and that spectral and geographical information are both used together.