Integrated data analysis for mineral exploration A case study of clustering satellite imagery, airborne gamma-ray, and regional geochemical data suites
- Partitioning cluster algorithms have proven to be powerful tools for data-driven integration of large geoscientific databases. We used fuzzy Gustafson-Kessel cluster analysis to integrate Landsat imagery,. airborne radiometric, and regional geochemical data to aid in the interpretation of a multimethod database. The survey area extends over 3700 km(2) and is located in the Northern Cape Province, South Africa. We carefully selected five variables for cluster analysis to avoid the clustering results being dominated by spatially high-correlated data sets that were present in our database. Unlike other, more popular cluster algorithms, such as k-means or fuzzy c-means, the Gustafson-Kessel algorithm requires no preclustering data processing, such as scaling or adjustment of histographic data distributions. The outcome of cluster analysis was a classified map that delineates prominent near-to-surface structures. To add value to the classified map, we compared the detected structures to mapped geology and additional geophysicalPartitioning cluster algorithms have proven to be powerful tools for data-driven integration of large geoscientific databases. We used fuzzy Gustafson-Kessel cluster analysis to integrate Landsat imagery,. airborne radiometric, and regional geochemical data to aid in the interpretation of a multimethod database. The survey area extends over 3700 km(2) and is located in the Northern Cape Province, South Africa. We carefully selected five variables for cluster analysis to avoid the clustering results being dominated by spatially high-correlated data sets that were present in our database. Unlike other, more popular cluster algorithms, such as k-means or fuzzy c-means, the Gustafson-Kessel algorithm requires no preclustering data processing, such as scaling or adjustment of histographic data distributions. The outcome of cluster analysis was a classified map that delineates prominent near-to-surface structures. To add value to the classified map, we compared the detected structures to mapped geology and additional geophysical ground-truthing data. We were able to associate the structures detected by cluster analysis to geophysical and geological information thus obtaining a pseudolithology map. The latter outlined an area with increased mineral potential where manganese mineralization, i.e., psilomelane, had been located.…
