Abstract
Understanding of the lateritic regolith units is important for regional geological and pedological mapping and provides valuable insights into landscape evolution and mineral exploration, especially in tropical environments. To contribute to these issues, this study used Landsat-8 OLI and Sentinel-2 MSI imagery to map the different lateritic regolith units in an area of 133 thousand km2 in the Midwest of Brazil. The approach used the Directed Principal Component Analysis (DPCA) technique and SRTM data as predictor variables for the Random Forest (RF) classification model, with the aim of comparing the efficiency of both sensors following identical classification methods. The results showed a similar overall accuracy of 73% for Landsat-8 and 74% for Sentinel-2, however the Sentinel-2-based maps are the most accurate with a kappa of 0.63 and a tau of 0.68. The significant levels of accuracy highlight the efficiency and feasibility of the multispectral imagery and machine learning approach to lateritic regolith mapping. It also emphasizes the importance of considering the spectral and spatial characteristics of sensors when selecting the most appropriate one for geological, pedological and geomorphological studies in tropical regions.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call