An Improved AI-Enabled Lithological Mapping Technique of Rocks using Multispectral Satellite Data
Year: 2025
Type: preprint
Abstract: Lithological mapping is vital for geological interpretation, mineral exploration, and land-use planning. This study develops an automated framework for classifying three key lithologies—Quartzite, Granite, and Carbonate—across diverse Indian formations using multispectral satellite data (ASTER and Landsat 8/9) combined with SRTM-derived terrain attributes. Supervised machine learning algorithms, Support Vector Machine (SVM) and Random Forest (RF), were applied to evaluate lithological discrimination across both datasets. Validation against reference geological maps showed that SVM consistently outperformed Random Forest, with Landsat-based SVM achieving the highest overall accuracy (91.2%, κ = 0.89) and ASTER-based SVM slightly lower (90.1%, κ = 0.87). Random Forest delivered competitive results (Landsat: 89.1%, κ = 0.85; ASTER: 88.3%, κ = 0.84). Confusion matrix analysis revealed that Quartzite and Carbonate occasionally overlapped spectrally, whereas Granite was more reliably classified. These results demonstrate that integrating multispectral data with machine learning enables robust lithological mapping. The approach highlights the scalability of SVM for semi-automated lithological classification in regions with limited field accessibility.
Authors: Tanushree Parmar, Romit Beed
Institution: St. Xavier's College
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W4413826775