Vegetation and land use/land cover of the Altar Valley, AZ mapped from Sentinel-2 data (2024)

This data release provides 10-meter vegetation and land-use/land-cover (LULC) maps for the Santa Margarita Ranch and the surrounding Altar Valley and Brawley Wash region of southern Arizona. Maps were generated for 2024 using two complementary Random Forest classification approaches implemented in Google Earth Engine (GEE): (1) a seasonal Sentinel-2 (S2) spectral–topographic model based on early- and late-season (ESLS) composites, and an additional model (2) based on Google’s Satellite Embedding V1 dataset. Both approaches incorporate 21 LULC classes derived from the USGS GAP CONUS 2011 land cover classification scheme available in GEE. Developed and agricultural classes were adapted directly from the 2021 National Land Cover Database (NLCD) to ensure consistent representation of human-modified land. Across the two models, predictor variables include Sentinel-2 surface reflectance bands, vegetation and moisture indices, seasonal differences, Google embedding bands, and 10-m terrain derivatives from the USGS 3D Elevation Program (3DEP). Each classification product is provided as a 10-m GeoTIFF accompanied by an ArcGIS auxiliary metadata file (.tif.aux.xml) that stores the raster’s virtual attribute table (VAT). The VAT contains the GAP land-cover codes, class descriptions, and color scheme used for visualization, along with the corresponding NLCD 2021 class codes, descriptions, and colors. NLCD attributes are included to document pixels where developed and agricultural classes were burned into the final map and to provide a built-in crosswalk, allowing users to reclassify the dataset into the NLCD scheme if needed. Accuracy assessment results are provided as CSV files and include overall accuracy, Kappa, per-class producer’s and user’s accuracy, and the confusion matrix.