Joint processing of Sentinel-2 and PRISMA data for monitoring urban areas (TUS:CAN Project)

The classification of man-made materials using multispectral and hyperspectral satellite data plays a key role in supporting sustainable urban planning and land management. In this context, the TUS:CAN project, carried out under the Italian Space Agency (ASI) program ’Innovation for Downstream Preparation - Public Administrations’, in collaboration with the Tuscany Region and the University of Siena, introduced a modular workflow for detailed mapping of artificial surface cover in urban settings. Level-2A Sentinel-2 images were hypersharpened to enhance the spatial resolution of the 20 m VNIR and SWIR bands to the 10-m native spatial resolution of the B2, B3, B4, B8 bands. The enhanced data were then processed using a Random Forest classifier to generate a binary artificial/natural land cover map. In parallel, PRISMA hyperspectral data were processed through geometric co-registration and pansharpening of the VNIR-SWIR bands using the mission’s panchromatic band at 5m spatial resolution. Independent artificial/natural cover maps were derived from each data source and subsequently fused to enhance spatial consistency and classification robustness in different urban forms. A further advancement involved the detection of less common artificial materials, such as metals and synthetic surfaces. These materials, among others, were identified using a matched-filter algorithm applied to the PRISMA hyperspectral cube. Ground-truth spectral signatures for common and rare materials were collected with an ASD FieldSpec 3 spectroradiometer and used for material classification and validation. Collectively, the integration of multispectral and hyperspectral data, map fusion techniques, and rare material detection represented significant progress towards scalable operational remote monitoring of urban materials.