Application of Machine Learning Methods for Hydrothermal Alteration Zoning using Remote Sensing Data: a Case Study of the Koldar Massif

Abstract. Hydrothermal alteration zones are among the key indicators in the exploration of porphyry copper deposits. In this study, a remote sensing-based approach was implemented to map hydrothermal alteration zones using ASTER satellite data and built-in classification algorithms available in the ENVI software environment. The study area is the Koldar massif, located in southeastern Kazakhstan within the Balkhash–Ili metallogenic belt, known for its intense hydrothermal alteration processes.
Four classification methods were applied: Spectral Angle Mapper (SAM), Support Vector Machine (SVM), Maxi-mum Likelihood (ML), and Minimum Distance (MD). The training samples were generated based on geological maps, lithogeochemical data, and expert visual interpretation. The focus was placed on mapping four types of alteration: argillic, phyllic, propylitic, and potassic zones.
Among the tested algorithms, the SVM method demonstrated the highest performance, achieving an overall classification accuracy of 84.12% and a kappa coefficient of 0.79. Propylitic and phyllic zones were effectively identified, while argillic and potassic zones showed partial spectral confusion due to similar spectral characteristics. The resulting maps show good agreement with geological structures and known mineralized zones of the Koldar intrusion, confirming the applicability of the chosen approach at early stages of exploration in arid environments. This approach provides a reproducible framework for mapping hydrothermal alteration zones and can be adapted for other porphyry systems using medium-resolution multispectral satellite data.