Northern Norway has an extensive geological variation compared to other parts of Norway. It is home to major ore deposits containing base and precious metals. It is located within Fennoscandia, which has been identified as a significant metal-producing region in Europe. Recently, Northern Norway has extensively been surveyed using several geophysical techniques, which led to coverage improvement of basic geological information relevant to the assessment of its mineral potential.
The study area, in the vicinity of Karasjok, is located on the Karasjok Greenstone Belt and comprises mainly Palaeoproterozoic rocks. This terrane is located on the same Palaeoproterozoic greenstone terrane of northern Finland, in which the detected gold deposits showed, with a few exceptions, similar characteristics to gold occurrences in Palaeoproterozoic greenstone belts in other parts of the world. Also, economic mineral deposits, e.g. Ni-Cu-PGE-Au mineralization, are largely restricted to events that occurred during the Palaeoproterozoic era. However, the study area for this study still remains largely under-explored and previous research and field work of the Karasjok Greenstone Belt have proven unsuccessful to locate significant mineral deposits. This area contains very few outcrops and traditional methods to map such areas are challenging.
A large number of different geophysical data sets exist for Norway. While individual data sets cannot provide direct information about the lithology of the mapped area, a combination of multiple data sets makes it possible to delineate and interpret characteristic regions of potentially similar lithology. This is of particular interest where the geological map may be unreliable or that geological mapping itself is difficult, for example, due to difficult accessibility or large vegetation coverage.
In this study, three fuzzy clustering algorithms, i.e. the fuzzy c-means algorithm, the Gustafson-Kessel algorithm, and the possibilistic fuzzy c-means algorithm, have been applied to compile pseudo-lithology maps and to characterize zones of mineralization by integrating recently acquired high-resolution airborne geophysical data sets of Karasjok. These algorithms have been applied for the integration of long-wavelength data sets, i.e. gravity and magnetic data reflecting deep geological features, the integration of short-wavelength data sets, i.e. gravity and magnetic data reflecting shallow geological features, and the integration of three near-surface data sets, i.e. the two short-wavelength data sets and a potassium data set. Analyzing the interrelationship of the data sets and pre-processing the data sets, have shown to be necessary steps in order to determine what kind of geological differentiation of the study area will be obtained when applying the clustering algorithms.
The application of the fuzzy clustering algorithms for the compilation of pseudo-lithology maps have shown promising results. While the long-wavelength data sets indicated different, likely deeper, geological features than has been mapped on the current bedrock map and did not show similar structures on a lo- cal scale, the clustering with short-wavelength data sets indicated similarities on a local scale. Adding the potassium data set as a third dimension, results in patterns that clearly reflect the presence of the Archaean tonalities gneisses and granites of the Jergul Gneiss Complex within the study area. This showed how the use of clustering with three data sets can obtain a larger geological differentiation of the study area. It is concluded that the local structures have a more detailed composition than currently mapped.
The application of clustering algorithms for the identification of zones of mineralization for this study has proven to be a powerful tool as it allows for the extraction of clusters of interests. One can simply use the clusters of interest to outline the areas of interest. In this case, the focus was to find areas that contain possible Ni-Cu-PGE mineralization, which are commonly reflected by high Bouguer and high magnetic anomalies. During this study, the clustering algorithm not only identified zones that were mentioned by previous re- search to contain possible mineralization, but also three additional zones that have not been explored or identified before.