Identification and mapping of Algerian island vegetation using high-resolution images (Pléiades and SPOT 6/7) and random forest modeling.

Despite their proximity to the coast, few studies have focused on identifying and mapping the vegetation of Algerian islands and islets. To fill this lacuna, our work, using satellite images and machine learning methods, is mainly aimed at identifying and mapping the main vegetation groups on a few islands, while evaluating the effectiveness of the random forest classifier, which is effectively used in the study of the vegetation of large areas. However, despite the high heterogeneity of their vegetation cover, the use of very high-resolution images (Pléaides and SPOT 6/7), through the fusion bands and derived bands (NDVI), has allowed the elaboration of a fairly precise vegetation map that can be used for the preparation of management and protection plans for these habitats. Our methodological approach revealed very satisfactory results, having allowed the identification of the plant communities inventoried in the field, while showing high accuracy values, ranging from 0.642 for the halophilic group of Asteriscus to 1 for the endemic Chasmophyte group of the Habibas archipelago (Pléiades images). The groups identified from SPOT 6/7 images show accuracy values between 0.67 for the Mediterranean cliff formations on Garlic Islet and 1 for the two formations (shrubby and herbaceous) of the Skikda islands. Our methodological approach, and notwithstanding the great heterogeneity and the very small surface areas of our islands and islets, has led to very satisfactory results, reflected with good overall accuracy and kappa index values (for Pléiades: overall accuracy > 92% and kappa index > 0.90; for SPOT 6/7: overall accuracy > 83% and kappa index > 0.80).