Contribution of artificial intelligence to the geological mapping of the SILET region (Western Hoggar) using aero-geophysical and satellite data

Abstract

Geological mapping is a fundamental task in the study of the Earth's crust, as it provides crucial insights into the structure, composition, and evolution of the planet's surface. Traditionally, geological mapping has relied on surface observations, geological drilling, and other time-consuming and expensive techniques. However, in recent years geophysical data has emerged as a valuable tool for enhancing geological mapping, allowing for more efficient and accurate characterization of the subsurface. This thesis explores the application of machine learning techniques to geophysical and satellite data for geological mapping. Specifically, we focus on the integration of airborne magnetic and gamma ray spectrometry data with Landsat images of the Silet region located in central Hoggar. Our goal is to improve our understanding of the geology of this region and explore the effectiveness of machine learning algorithms in this context. Our findings show that geophysical data can provide valuable information on the subsurface structure and lithology, which can help to refine geological interpretations and reduce uncertainty in geological maps. In particular, we demonstrate the importance of integrating geophysical data with geological observations, as well as the importance of high-quality data acquisition and processing. Additionally, we show that machine learning techniques can help to automate the interpretation of geophysical data and improve the accuracy of geological maps. In our case study, we applied a range of machine learning algorithms, including random forests (RF), Deep neural networks DNN) and extreme gradient boosting (XGBoost). We demonstrate that these algorithms can effectively classify geophysical data into different lithological units and identify subsurface structures. Specifically, we show that the machine learning tool can distinguish different rock types and identify the boundaries between different rock units based on magnetic and gamma ray spectrometry data. Overall, this thesis provides a comprehensive overview of the contribution of machine learning applied to geophysical data for geological mapping, and highlights the potential for the utility of these data to revolutionize our understanding of the Earth's crust