Publications Repository - Helmholtz-Zentrum Dresden-Rossendorf

Switch on selection | Login

1 Publication
Multi-source hyperspectral imaging of carbonatite-hosted REE-Nb-Ta mineralization at Marinkas Quellen, Namibia
Booysen, R.; Zimmermann, R.; Lorenz, S.; Gloaguen, R.; Nex, P. A. M.;
The demand for mineral and metalliferous resources needs to match the continued global rise in population and global economic growth. Rare Earth Elements (REEs), Niobium (Nb) and Tantalum (Ta) are such deposits in high demand. This global rise makes it difficult to meet the growing demand using only the currently available resources, such as recycled REEs and known REE deposits. Although the concept of a purely circular economy is very attractive through the use of recyclable REE-Nb-Ta, this model is not completely sustainable due to the increased energy needed to bolster such a model. Therefore, a renewed focus on the exploration of REE-Nb-Ta deposits is imperative to ensure the future development of this commodity.
Traditional exploration techniques are mainly based on extensive field work supported by geophysical surveying. Restrictions such as field accessibility, financial status, area size and climate can hinder these traditional exploration techniques. Hence, we suggest to increase the use of multi-source and multi-scale hyperspectral remote sensing in order to decrease conventional restrictions in the exploration of minerals through the use of aerial and ground-based methods. The multi-scale, multi-source approach will consist of a downscaling procedure, moving from low spatial resolution to high spatial resolution. Firstly, satellite data (Sentinel-2) will be used to identify the study area, then hyperspectral airborne data (HyMap) will be used to refine the area of interest. Subsequently, a snapshot hyperspectral camera will be attached to a UAV to acquire drone-borne data for the investigation of the deposit in more detail. We further argue that the addition of drone-borne hyperspectral data can also vastly improve the accuracy of field mapping in future mineral exploration. Drone-borne measurements can supplement and direct geological observation immediately in the field and therefore allow better integration with in-situ ground investigations. In particular, in inaccessible and remote areas with little infra-structure, such systems are an excellent reconnaissance tool because it allows a systematic, dense and completely non-invasive surveying, which is often not possible using ground-based techniques. Additionally, spectral and spatial information will be integrated by combining drone-borne hyperspectral and Light Detection And Raging (LiDAR) data to provide more accurate classification results.
Ultimately, the corrected drone-borne data provide information on the spectral signatures of outcropping lithologies to the exploration teams. This is achieved by using end-member modelling and classification techniques such as non-linear machine learning algorithms, e.g., Neural Networks and decision tree based methods. The drone based data are integrated in a comprehensive workflow including in-situ acquisitions and results in an hypercloud. The validation of the resulting digital outcrop is performed via field spectroscopy, portable XRF and representative geochemical whole-rock analysis.
The area of interest for this study is the massive carbonatite intrusion at Marinkas Quellen, Namibia. The location is in a remote environment and characterized by difficult terrains and a complete carbonatite suite (e.g. calsio-, ferro- and magnesio-carbonatites). The first two factors would normally impede or restrict traditional field surveying. Preliminary results indicate that drone-borne surveying has a very high potential to directly detect REE-concentrations and indicator minerals for Nb and Ta, in fundamentally lowering the acquisition costs and increasing the information potential of data captured in the field.
Keywords: REEs; Multis-source; Hyperspectral; Exploration; Marinkas Quellen
  • Poster
    WHISPERS - Hyperspectral Image and Signal Processing Workshop, 23.-26.09.2018, Amsterdam, The Netherlands

Publ.-Id: 28552 - Permalink