Uso de Sentinel-2 y datos auxiliares para la generación, mediante clasificación de imágenes, del Mapa de Usos y Cubiertas del Suelo de Cataluña 2017

Abstract

<p>This paper details the process of generating the 2017 Land Use/Land Cover Map of Catalonia (MUCSC) using automatic classification of satellite imagery and auxiliary cartographic and remote sensing data. A total of 60 images (6 dates for each of the 10 tiles covering Catalonia) captured by the Sentininel-2A and Sentinel-2B satellites were used. These images as well as texture variables, terrain models derived from lidar processing, and vegetation and wetness indices were classified using the k-Nearest Neighbor algorithm (kNN) to obtain a map with 25 categories. The categories related to urbanized areas (urban areas, urbanizations and industrial zones/ commercial areas), road infrastructures and burned areas were edited using official cartographic datasets of the Catalan Government [Generalitat]. The results have an overall accuracy greater than 98 %, which was evaluated with a set of more than 8.6 million independent test pixels. This work represents an important milestone in terms of the computational effort it involves due to the territorial extension (32 000 km²), the spatial detail of between 2 and 20 m, the use of up to 58 variables, the relative completeness of the legend and the level of success achieved. The MUCSC 2017, which is part of a 30-year quinquennial series beginning in 1987, can be downloaded in different formats (also in MMZX: new ISO 19165-2) and at resolutions of 10 m and 30 m pixel side from the Ministry of Territory and Sustainability website of the Catalan Government.</p>