Critical Analysis of Co-Occurrence Measures for Landscape Characterization with High Resolution Satellite Data

A comprehensive evaluation of textural feature extraction methods, with a focus on second-order co-occurrence metrics derived from Gray-Level Co-occurrence Matrices (GLCM). Applied to high-resolution satellite imagery, these metrics reveal spatial variability but are sensitive to pixel size and terrain complexity. This study focuses on a $100 \mathrm{~m} \times 100 \mathrm{~m}$ area near the Solani River in Roorkee, India, chosen for its pronounced topographical variation. Using GLCM-derived features:- mean, variance, homogeneity, contrast, dissimilarity, entropy, second moment, and correlation across four spectral bands, it produces 36 detailed texture layers. The analysis assesses these metrics’ effectiveness in characterizing terrain with depth changes over 4 meters, supplemented by spatial metrics like lacunarity, Moran’s I, and Geary’s C to capture clustering and pattern distribution. Ground-truth validation data from show over 76% accuracy, with contrast and entropy as strong indicators of terrain shifts. While limited by GLCM sensitivity and spatial resolution, texture-based remote sensing effectively complements spectral analysis, uncovering subtle landscape features.