This paper presents a framework for combining airborne LiDAR and hyperspectral PRISMA data to create a more comprehensive input for the Machine Learning (ML) models. To predict Cobalt concentration, we apply three well-known ML methods: Random Forest (RF), Support Vector Regression (SVR), and Multi-Layer Perceptron (MLP). A key challenge in this application is the limited availability of labeled data, which we address by employing three data augmentation techniques, ranging from traditional methods to deep learning-based approaches, to generate synthetic data points. Experiments were conducted on a mineralization site at the Áramo mine in Asturias, Spain. The results demonstrate that these data augmentation techniques significantly enhance the ML models’ ability to accurately predict the minority class, which is crucial for mineral exploration. Combining data from LiDAR and PRISMA improves model performance compared to using a single modality.