Tree Species Identification Based on Phenological Trajectory Divergence in Time Series

We present a tree species identification method based on phenological trajectory divergence in time series (PTDTS), utilizing exclusively open-access datasets (Sentinel-1/2 and NASADEM). The data were clipped and processed according to the vector boundaries of our study area. Due to the confidentiality constraints associated with mining regions (our study area), the specific boundary data cannot be disclosed. Therefore, this study primarily emphasizes the replicability of the methodology. Tree species identification based on PTDTS primarily incorporates four analytical dimensions: time-domain features, spatial features, dynamic features, and frequency-domain features. We process Sentinel-2 data to compute Normalized Difference Vegetation Index (NDVI) time series, then derive four spatiotemporal feature datasets through PTDTS analysis. These feature matrices serve as discriminative inputs for the tree species classification model.