Machine Learning and Remote Sensing Based Classification and Vegetation Dynamics Assessment of Inselberg Habitats in Gampaha, Sri Lanka

Abstract

Inselbergs are isolated rocky outcrops of ecological and geomorphological significance, yet their vegetation dynamics remain understudied in Sri Lanka. This study applied remote sensing and machine learning techniques to classify inselberg habitats in the Gampaha District and to assess vegetation changes over time. A Support Vector Machine (SVM) classifier was developed using selected topographic and spectral covariates, achieving an overall accuracy of 94.5% and a Kappa coefficient of 70.1%. The Digital Elevation Model (DEM), Ferrous Mineral Ratio (FMR), and Sentinel-2 Band 8 (NIR) were identified as the most influential predictors of inselberg presence. Vegetation change was evaluated using NDVI-based change detection between 2001 and 2024, and results indicate notable vegetation reduction in several inselberg habitats. Quarry expansion emerged as the dominant anthropogenic driver of degradation, while inselbergs associated with Buddhist monasteries exhibited comparatively higher vegetation retention. The study highlights the value of geospatial modelling for inselberg habitat assessment and provides insights that can guide targeted conservation planning in rapidly urbanizing landscapes.