Monitoring Posidonia oceanica meadows in a Mediterranean coastal lagoon (Stagnone, Italy) by means of neural network and ISODATA classification methods

ISODATA (Iterative Self-Organizing Data Technique of Analysis) and neural network classification methods were carried out to map shallow Posidonia oceanica meadows in coastal areas of the Mediterranean Sea, using Coastal Zone Colour Scanner (CZCS) airborne sensor data obtained at different altitudes and an aerophotogrammetric image. Reference test points of P. oceanica have been checked against aerial photographs. The neural-based classification method gives the best performance (92-95%) for all the images of the set, except for the highest altitude flight (1000 m, accuracy 74%). ISODATA classification of CZCS images was generally more accurate (81-85%) than applied to the aerophotogrammetric image (79%). The study also indicated that 4 m represents the 'critical' resolution useful for the extraction of reliable information within the study analysed area. Where P. oceanica forms dense and continuous meadows, a lower resolution (such as those obtainable from satellite sensors) could be successfully applied.