Temporal aggregation strategies for urban vegetation and tree species classification using PlanetScope imagery - data

The raw dataset used in this study integrates high-resolution, multi-temporal PlanetScope imagery with field-derived reference data for both major vegetation classes and tree species. PlanetScope surface reflectance (Level-3B AnalyticMS) imagery with 3 m spatial resolution and four spectral bands (blue, green, red, and near-infrared) was acquired under Planet’s Education and Research Program, ensuring dense temporal coverage suitable for urban environments characterized by fragmented vegetation and rapid phenological variability. Monthly composites were generated from at least five cloud-free scenes per month for the period 2021–2023, resulting in 36 temporally consistent images used to derive spectral bands and vegetation indices (NDVI, GNDVI, MSAVI, EVI, and NIRv). Reference data consisted of two hierarchical levels: (1) major vegetation classes, including trees, shrubs, grass, and non-vegetation, and (2) species-level tree samples representing seven dominant urban tree species identified through multi-season field campaigns conducted in 2025. In total, 6,228 labelled samples were used for major class classification and 265 field-verified points for tree species classification, ensuring robust representation across classes. All spectral and index features were extracted and normalized to ensure comparability across time and space and were linked to the corresponding class labels and temporal attributes (month and year). This integrated dataset forms the basis for supervised classification, cross-validation, and subsequent analysis of urban vegetation structure and dynamics.