Comparative evaluation of fast-learning classification algorithms for urban forest tree species identification using EO-1 hyperion hyperspectral imagery

Introduction Accurate identification of forest tree species is essential for sustainable forest management, biodiversity assessment, and environmental monitoring. Urban forests, in particular, present spectral heterogeneity that challenges conventional classification methods. This study focuses on developing an efficient classification framework for species-level tree mapping in the Hauz Khas Urban Forest, New Delhi, India, using EO-1 Hyperion hyperspectral imagery. Methods Thirteen supervised classification algorithms were comparatively evaluated, encompassing traditional spectral/statistical classifiers—Maximum Likelihood, Mahalanobis Distance, Minimum Distance, Parallelepiped, Spectral Angle Mapper (SAM), Spectral Information Divergence (SID), and Binary Encoding—and machine learning algorithms including Decision Tree (DT), K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Random Forest (RF), and Artificial Neural Network (ANN). Dimensionality-reduction techniques (Principal Component Analysis—PCA and Minimum Noise Fraction—MNF) and band-selection strategies based on the Average Pairwise Absolute Difference (APAD) metric and species-specific band-ratio indices were implemented to mitigate spectral redundancy. Ground-truth samples were collected from extensive field surveys and validated using very high-resolution Pléiades imagery. Results A total of 21 tree species were identified. Among all classifiers, Random Forest and Decision Tree exhibited superior performance, with Random Forest achieving the highest species-level accuracy (95% for Peepal and Medlar) and overall accuracy of 82.56% (Kappa = 0.81) when applied to PCA-transformed data. Discussion The results highlight that integrating dimensionality reduction and optimal band selection with ensemble learning substantially improves classification efficiency and accuracy. The study identifies the most effective fast-learning classifiers for hyperspectral urban forest mapping and underscores the potential of hyperspectral imaging and ensemble methods for scalable and operational tree species monitoring.