Prediction of Ore Objects of Equatorial Africa Using Space Images

Automated forecasting of deposits based on the Data Mining tools implies the broadest use of heterogeneous and multi-level data – the results of geophysical, geological, landscape surveys and space imagery, among which namely the last are gaining ever more important role [1, 2].

Works were carried out on a site with the area of 2500 km2 on the territory of the Democratic Republic of the Congo and the Rwanda, near the Lake Kivu. The eastern and central parts of the site are located within the Great Rift Valley; the west is outskirts of the Congo Basin. The site is composed of rocks of Paleo- and Mesoproterozoic age, with impregnations of Neoproterozoic rocks and is partially overlapped by Neogene sediments. Within the site there are several tens of known deposits of tin, iron, niobium, tantalum, tungsten and other metals [3].

Because getting of high-quality geological and geophysical data had proved to be very problematic, the work was carried out only on the basis of space images.

Initial data were Landsat-8 multispectral images (the date of acquisition is July-September 2014), Google high-precision images and SRTM data. Work stages are as follows.

First, the raw data were preprocessed. On the one hand, maps of various spectral characteristics as band combinations (including spectral indices, such as NDVI and others) were created, followed by combining them with digital elevation models.

On the other hand, procedures of lineaments, circular and arc structures extraction of different sizes and degrees of fracturing were carried out. Such elements, in turn, served as the basis for calculation and construction special maps (density of lineaments with different azimuths, concentration of intersection points of multidirectional lineaments and others).

Both groups of materials were analyzed in terms of existence of spatial associations with known deposits.

At the next stage the Data Mining procedures – classification, multidimensional scaling, pattern recognition, ranking and others, were applied to initial and transformed data. Further these materials were compared with data of the previous stage. The results were presented in the form of complex predictive 3D maps with allocated on them prospective areas for each mineral.

Authors for the first time in their practice carried out forecasting of ore objects over a large area entirely using satellite images. The results proved to be surprisingly satisfactory enough.


1. Gornyy V.I., Tronin A.A. Review of the Last Decade Major Achievements of Remote Sensing Methods Application on the Geological & Geophysical Problems Solution // Current Problems in Remote Sensing if the Earth From Space. – 2012, Vol.9, №5. – pp. 116-132
2. Busygin B.S., Nikulin S.L. Integrated Analysis of Geological-Geophysical Data and Satellite Images for the Ore-Gold Mineralization Forecast // Current Problems in Remote Sensing if the Earth From Space. – 2009, Vol.6, №1. – pp. 17-23
3. Maarten de Wit, Francois Guillocheau, Michiel C.J. de Wit. Geology and Resource Potential of the Congo Basin. Regional Geology Reviews. – Springer, 2015. – 417 pp.