Our Impact

Deep Learning Approaches to Mineral Prospect Modeling of Rare Earths in Carbonatites

Thematic Area: Energy including renewables

University: University of Nairobi (UON)

Project Leader: Prof Hudson Angeyo Kalambuka

Collaborating Partners: Nelson Mandela African Institute of Science & Technology ; Dar es Salaam Institute of Technology

Duration: 2 years

Project Overview


There is increasing demand for innovative mineral prospecting to identify and define strategic deposits as well as delimit the geologic structures that host the deposits. Due to their unique properties many countries consider rare earth elements (REE) as a strategic resource for applications in advanced technologies and carbon-free based fuels. As mineralogical constraints exert significant control over the extraction of rare earth elements from their ores, the project seeks to develop a multiplexed spectroscopy and imaging analytical protocol leveraged by deep learning, for prospective modelling of REE occurrence in selected carbonatite complexes of East Africa relative to their genesis and geology.

The technology

The REE prospective methods currently used are mostly linear and parametric, and so cannot accurately model the multivariate processes associated with geochemical distributions of the mineral resources and that aggregates the complex attributes of carbonatite REE potential.  Deep learning will be used to develop multivariate calibration strategies for both spectral and imaging analyses as well as to model geochemical distributions and to extract associations related to REE occurrence. This will help to define the geologic footprint of REE carbonatite deposits, and to extrapolate this knowledge to areas not well characterized to help develop a predictive exploration model for the identification of REE-bearing deposits in various carbonatites.

Expected Impact

The project will greatly contribute to sustainability of East Africa’s strategic mineral and energy resources


For more information, please contact the Project Leader