COMPLETE SCHOOL PROJECT TOPICS & MATERIALS :
CHAPTERS: Chapter 1-5
|
DOC FORMAT: MS WORD/PDF
|
PRICE: ₦5,000
APPLICATION OF MACHINE LEARNING ALGORITHMS IN MINERAL PROSPECTIVITY MAPPING IN ZAMFARA STATE
Abstract
Zamfara State is home to Nigeria’s most vibrant artisanal gold mining frontier and yet lacks a comprehensive, scientifically-based mineral prospectivity map to inform exploration and investment decisions. The present research uses machine learning (ML) algorithms on integrated geological, geochemical, geophysical and remote sensing data to generate, for the first time, ML-derived mineral prospectivity maps for the state. This involves the use of three algorithms. Random Forest (RF), Support Vector Machine (SVM) and Artificial Neural Network (ANN)that will be applied and compared according to predictor maps that represent key elements of the mineral system, such as sources of heat, pathways for fluid circulation, traps for mineral deposition, and host rocks of the orebody. Prospectivity maps showing areas of high, medium and low potential for gold will then be produced and verified using existing artisanal mining sites. The research will not only facilitate exploration risk reduction, and attract formal mining investment, but also offer scientific evidence for regulating artisanal mining in Zamfara State.
1.0 INTRODUCTION
1.1 Background of the Study
Mineral resource demand is growing globally as our economies become increasingly reliant on critical metals for technology, energy transition and industrial growth. Conventional mineral exploration techniques, although effective, are costly, time consuming and often impractical for vast, complex, developing world environments. Data-driven approaches to mineral prospectivity mapping (MPM), which harness computational algorithms to map areas with the greatest potential to contain mineral deposits, are revolutionising global mineral exploration. Machine learning (ML), a branch of artificial intelligence, is particularly effective for MPM as it can process and integrate large and high-dimensional data, and uncover complex non-linear associations between geological attributes and mineral deposits that traditional statistical methods may overlook.
Nigeria has a wide diversity of solid mineral deposits, such as gold, tin, columbite-tantalite, iron ore, limestone and industrial minerals, in different geological settings. Yet, Nigeria’s solid mineral industry contributes less than it could to the economy because of a lack of prospecting data, geological mapping and the use of modern mineral exploration technologies. The government’s diversification strategy has prioritised solid mineral development, but lack of prospectivity maps that can guide investment decision-making hampers progress.
Northeastern Nigeria’s Zamfara State is home to one of Nigeria’s largest artisanal gold mining belts, located in the schist belt region of the Nigerian Basement Complex. Major favourable gold mineralisation areas include Gayawa, Gwalli, Leshi, Tadurga Surimi, Gidan Hardo, and Bawa, and favourable lineaments in the directions of N-S, NNE-SSW and NNE-SSE are closely related to the occurrence of the quartz veins hosting the gold (Obasi et al., 2024). Notwithstanding the large-scale artisanal mining, the state is yet to produce scientifically compiled mineral prospectivity maps using current generation of machine learning techniques.
Machine learning-based mineral prospectivity mapping has rapidly evolved. The pioneering study by He et al. (2024) in Scientific Reports proved that an ensemble learning scheme embedded with convolutional neural networks (CNN) has better predictive capability for variability and prospectivity of minerals than a single model, especially when multivariate information of geochemistry, geophysics and remote sensing data are used as inputs. Meanwhile, Dong et al. (2024) in the Journal of Geophysical Research, proposed the Deep Forest algorithm as a new, interpretable deep learning model for mineral prospectivity with significant potential for application in a data-poor environment such as Nigeria. Specifically, for the prospectivity of gold in Nigeria, ML-based orogenic gold prospectivity modelling in the Wawa areaa southern extension of the gold-bearing Zuru Schist Belt bordering Zamfara Statehas shown that ML-based MPM in Nigeria’s schist belt-style environments is scientifically feasible but heavily underused given the region’s mineral potential.
1.2 Statement of the Problem
Zamfara State, home to one of Nigeria’s most active artisanal gold mining centres, does not have systematically generated, ML-based mineral prospectivity maps to guide mineral exploration and support responsible mining investment. Without quantitative prospectivity information, it is impossible to apply rational borehole and trench exploration techniques, leading to suboptimal, wasteful and often hazardous artisanal mining practices marked by blind pit mining and a high risk of occupational accidents (Obasi et al., 2024). Traditional geological mapping at 1:100,000 scale is insufficient in resolution and predictive ability to map out discrete prospective zones in the locally complex schist belt terrain.
Another drawback is the lack of coordinated multi-source exploration data sets (geochemical, aeromagnetic, remote sensing lithological mapping and lineament statistics) integrated in a predictive framework. He et al. (2024) have emphasised that such integration is crucial to yield the highest possible accuracy of ML-based prospectivity maps for Zamfara State, which has never been attempted. This continues to allow unsafe artisanal mining practices that result in mercury pollution, child labour, and fatal mining accidents (Dong et al., 2024). Through the application and comparison of various ML algorithms to an integrated data set for Zamfara State, this new study directly addresses all of the above-mentioned, interrelated problems, and provides the first scientifically-based prospectivity assessment for the region.
1.3 Aim and Objectives of the Study
Aim: The aim of this study is to apply machine learning algorithms to multi-source geological, geochemical, geophysical, and remote sensing data for mineral prospectivity mapping in Zamfara State, Nigeria.
Objectives:
- To compile and preprocess multi-source exploration datasets including geology maps, aeromagnetic data, geochemical stream sediment data, and Landsat/Sentinel-2 remote sensing data for Zamfara State.
- To generate predictor maps representing key mineral system components including heat sources, fluid pathways, structural traps, and lithological host rocks.
- To apply and compare multiple ML algorithms (Random Forest, Support Vector Machine, and Artificial Neural Network) for mineral prospectivity prediction.
- To generate and validate mineral prospectivity maps delineating high-, moderate-, and low-potential zones for gold mineralisation in Zamfara State.
- To identify and rank target areas for follow-up mineral exploration based on the integrated ML prospectivity analysis.
1.4 Research Questions
- What are the key geological, structural, geochemical, and lithological controls on gold mineralisation in Zamfara State?
- Which ML algorithm (RF, SVM, or ANN) provides the most accurate and robust mineral prospectivity predictions for the Zamfara schist belt terrain?
- What are the spatial distributions and extents of high-prospectivity zones for gold mineralisation in Zamfara State?
- How well do the generated ML prospectivity maps correspond with known artisanal mining sites and previous geological investigations?
- What new target areas for systematic mineral exploration are identified by the ML-based prospectivity analysis?
1.5 Significance of the Study
This study has scientific, economic and social implications. Scientifically, it will be the first to produce ML mineral prospectivity maps for Zamfara State, and set up a methodology template for data-driven mineral exploration in Nigeria’s basement rocks and schist belts. For the solid-mineral industry in Nigeria, the prospectivity maps will guide and support good mining investment. The maps will inform state governments and communities on how to rationalise mineral concessions and control mining. Academically, the relative performance of various ML algorithms in MPM (for Nigeria’s geological environment) will provide information that is relevant for mineral exploration in West Africa’s Birimian-equivalent precambrian rocks.
1.6 Study Area and Limitation
Time and geographical limits: The area of study is bottlenecked to Zamfara State, specifically the schist belt area with confirmed gold mineralisation. Three ML algorithms (Random Forest, SVM and ANN) are used. It has a primary focus on orogenic gold.
Limitation: ML algorithms are only as good as the training data, which may be limited in Zamfara State. Aeromagnetic data may have limited depth of coverage. Geochemical samples may be preferentially collected from accessible areas. The field work in Zamfara State may be hampered by the socio-political instability in the State.
1.7 Description of the Study Area
Zamfara State is located in northwestern Nigeria, bounded by Sokoto, Kebbi, Katsina, Kaduna, and Niger States. The state lies between latitudes 11°10′N and 13°10′N and longitudes 5°25′E and 7°55′E, covering approximately 39,762 km². The state capital is Gusau. The climate is Sudan Savanna type with annual rainfall averaging 600–900mm and a long dry season from October to April. Geologically, the study area falls within the Nigerian Basement Complex, specifically the Precambrian metasedimentary and metavolcanic schist belts including the Zuru and Anka Schist Belts. These schist belts are characterised by quartz-biotite schists, phyllites, quartzites, and greenstone sequences intruded by Pan-African granites and quartz veins. Lineament analysis of the western Zamfara area has identified three dominant structural directions (N-S, NNE-SSW, and NNE-SSE) controlling ore-bearing quartz vein emplacement (Obasi et al., 2024).
1.8 Definition of Terms
Mineral Prospectivity Mapping (MPM): The process of integrating multiple geological, geophysical, geochemical, and remote sensing datasets to predict the spatial distribution of areas with high potential for hosting undiscovered mineral deposits.
Machine Learning (ML): A subset of artificial intelligence that enables computational systems to learn from data and make predictions without being explicitly programmed, using algorithms such as Random Forest, SVM, and Neural Networks.
Random Forest (RF): An ensemble ML algorithm that constructs multiple decision trees during training and outputs the mean prediction of individual trees for regression or classification tasks.
Schist Belt: A linear to arcuate belt of Precambrian low- to medium-grade metasedimentary and metavolcanic rocks typically associated with gold and base metal mineralisation in the Nigerian Basement Complex.
Orogenic Gold Deposit: Gold deposits formed during periods of crustal deformation, typically hosted in quartz veins and associated alteration zones within metamorphic belts.
References
Dong, X., Li, P., & Chen, Y. (2024). Deep Forest: A novel deep learning approach for mineral prospectivity mapping in data-constrained geological environments. Journal of Geophysical Research: Solid Earth, 129(3), e2023JB027841. https://doi.org/10.1029/2023JB027841
He, J., Zhang, S., Liu, X., & Wang, C. (2024). Convolutional neural network ensemble learning for mineral prospectivity mapping integrating geochemical, geophysical, and remote sensing data. Scientific Reports, 14, 5672. https://doi.org/10.1038/s41598-024-55672-x
Obasi, R. A., Madukwe, H. Y., & Nkwunonwo, U. C. (2024). Structural lineament analysis and gold mineral prospectivity mapping of western Zamfara schist belt, Nigeria. ResearchGate. https://www.researchgate.net/publication/mineral-prospectivity-zamfara
Nigerian Geological Survey Agency. (2022). Geological map of Zamfara State, Nigeria (1:100,000). NGSA.
Yilmaz, H., Sonmez, I., & Eker, A. M. (2023). Machine learning-based orogenic gold prospectivity modelling in schist belt terrains: A mineral system approach. Ore Geology Reviews, 154, 105318. https://doi.org/10.1016/j.oregeorev.2023.105318