Lithological Mapping in High-Vegetation Areas Using Sentinel-2, Sentinel-1, and Digital Elevation Models

Area of Science:

• Geospatial analysis
• Remote sensing for geology
• Machine learning in earth science

Background:

• Lithological mapping is difficult in high-vegetation areas due to obscured surface features.
• Accurate geological information is crucial for understanding ecosystem limitations.

Purpose of the Study:

• To develop an effective method for lithological mapping in high-vegetation areas.
• To assess the utility of multi-source remote sensing data and machine learning for this task.

Main Methods:

• Utilized Sentinel-2, Sentinel-1, and Digital Elevation Model (DEM) data.
• Employed the Random Forest algorithm for supervised classification.
• Focused on a high-vegetation coverage area in Taiwan Province, China.

Main Results:

• The optimal combination of Sentinel-2 and DEM data achieved 84.30% overall accuracy and a Kappa coefficient of 0.83.
• Spectral features (B2, NDBI) and topographic features (elevation) were significant predictors.
• Demonstrated the effectiveness of the Random Forest model in this challenging environment.

Conclusions:

• Multi-source remote sensing data, particularly Sentinel-2 and DEM, combined with Random Forest, significantly improves lithological mapping accuracy in high-vegetation areas.
• Spectral and topographic features are key indicators for lithological identification.
• The findings offer valuable insights for geological surveys and ecosystem studies in similar environments.