Dados da Submissão

Título

Multispectral Survey Based on Remote Sensing for Chromite Prospecting in Oman Ophiolite

Texto do resumo

This study presents the results of a mineral exploration campaign in the Semail Ophiolite in Oman. The objective was to develop a predictive model for chromite exploration using remote sensing techniques, geoprocessing, and data science applied to high-resolution multispectral orbital images. The study area, located within the ophiolitic formation in Northern Oman, spans 104 km². The WorldView-3 satellite images, acquired by OCC, provide data in the Visible and Near-Infrared (VNIR), Shortwave Infrared (SWIR), and Panchromatic bands.
The primary objectives of this study were to utilize high-resolution multispectral orbital images to develop a predictive model for chromite exploration, apply Principal Component Analysis (PCA) to differentiate geological features and enhance lithological interpretation, employ various machine learning techniques to predict areas with high potential for chromite mineralization, integrate spectral data and machine learning models to create a condensed predictive model using band arithmetic, and validate the predictive model by identifying potential exploration targets and comparing them with known mineralization zones.
The methodology involved acquiring orbital reflectance data in various bands (VNIR and SWIR), which help distinguish materials based on their spectral signatures. The VNIR and SWIR bands were stacked to generate a 16-band image, providing detailed spectral information. Principal Component Analysis (PCA) was applied to differentiate geological features, enhancing the interpretation of lithological responses. Additionally, the SpecHunter software was used to capture spectral signatures of pixels, aiding in the identification of potential targets. Several machine learning techniques were applied to the data cube to predict areas with chromite mineralization potential (Support Vector Machine (SVM); Random Forest; Neural Networks; IsoCluster).
With the results of the models, we created a condensed model using band arithmetic, where pixels with more recurrence in the models change in color scale.
The integration of remote sensing and advanced data analysis has significantly enhanced the prediction of chromite mineralization zones in northern Oman. Machine learning models effectively identify potential exploration targets, aiding efficient resource allocation.
Additionally, PCA results highlighted discordant structures in duct shapes, aligning with Braun and Kelemen's (2002) model of dunite conduits formed by magmatic flows. The interaction of MORB-type magmas with mantle harzburgites led to magmatic flows and metasomatic processes, confirmed by mineral chemistry studies (Cocomazzi, 2020). These findings suggest a metallogenic relationship between dunite ducts, gabbro dikes, and chromite mineralization zones, providing a robust framework for future prospecting.

Palavras Chave

Orbital Data; WorldView 3; Machine Learning; Chromite