Sentinel-2 Level-2A Spectral Indices Dataset (2023) for Machine Learning-Based Soil and Vegetation Parameter Retrieval in the Trás-os-Montes Region, Northeastern Portugal

Data are organized into four seasonal periods (April, May, July, and November), each comprising 43 GeoTIFF tiles that together cover the full study area. A total of 70 spectral indices are included, spanning five thematic categories: vegetation (NDVI, EVI, EVI2, SAVI, MSAVI, OSAVI, TSAVI, SATVI, DVI, RDVI, RVI, GNDVI, GRVI, GLI, GARI, TVI, ARVI2, BWDRVI, IPVI, PANDVI, RB, MV, CCCI, ARI, TRIANDULARVI), soil (BI, BI2, BI3, BLI, BSI, GSI, NBR, NBR2, RI, SI, SI2, COLOR, FI, HUE, NDSI, CRSI), salinity (S1, S2, S3, S4, S5, S6, S7, S8, S9), water and moisture (NDWI, MNDWI, NDMI, LSWI, GVMI, MSI, S2WI, AWEI, V), and geology/mineralogy (CI, CLAY, Fe2, Fe3, FO, FS, GI, Gossan).

The dataset is designed to support machine learning and remote sensing applications requiring spatially explicit, multi-temporal spectral features for soil and vegetation parameter estimation, including soil organic carbon mapping, salinity detection, moisture assessment, and land cover classification. All indices were derived from Sentinel-2 Surface Reflectance (Level-2A) imagery with cloud masking applied prior to index computation.