Using very high resolution satellite imagery for land cover mapping in Pico Da Vara Nature Reserve (S. Miguel Island, Archipelago of the Azores, Portugal)

The effectiveness of very high-resolution IKONOS multispectral satellite imagery for land cover/vegetation mapping in Pico da Vara Nature Reserve (S. Miguel Island, Archipelago of the Azores, Portugal) was assessed. A per-pixel supervised classification scheme was developed using two different datasets. The Dataset A included the IKONOS-2 ortho-rectified and atmospherically corrected Visible and Near Infrared (VNIR) multispectral bands (4 bands in total). Besides the same spectral bands of Dataset A, Dataset B also included two additional GIS-based raster bands representing both `Altitude´ and `Distance to water streams´ layers, since these two environmental variables are strongly relevant for land cover/vegetation categories distribution in Pico da Vara Nature Reserve. Overall, most of the five supervised classification techniques applied (Support Vector Machine; Maximum Likelihood; K-Nearest Neighbor; Linear Discriminant Analysis; Probabilistic Neural Network) showed a strong agreement and a good accuracy. The best classification accuracy was obtained by applying the KNN classifier to Dataset B. Remote sensing by using very high-resolution IKONOS-2 multispectral imagery was confirmed as an effective method for land cover mapping in Pico da Vara Protected Area.