|dc.description.abstract||Despite groundwater being a very important and valuable source of water for humans it is under
serious threat by human activities such as mining. Nevertheless, benefits of mining, in particular coal
mining, have led to improvement in the social and economic development of many countries.
However, despite the important benefits of coal to developing and developed countries, groundwater
pollution caused by coal mining generated pollutants (acid mine drainage (AMD)) is a significant
problem globally. Finding scientific solutions on how coal mining and coal waste disposal can be
controlled and monitored from a policy and decision-making perspective in view of ultimately
minimising the pollution of groundwater resources in coalfields is therefore an urgent imperative.
This can be done by undertaking a coalfield-scale groundwater vulnerability assessment to rank areas
in terms of the likelihood of pollutants reaching the groundwater.
The research discusses the development of a new approach for groundwater vulnerability assessment
specific to AMD at a regional coalfield scale using artificial intelligence to automate some of the
processes and to assist in the generation of rapid groundwater vulnerability assessment results for use
by policy and decision-makers in controlling coal mining and coal waste disposal in a typical
coalfield. The approach is a hybrid of the index and overlay, process-based simulation and statistical
approaches generated by combining factors associated with pollutant sources, energy that drives the
pollution migration and the ways AMD pollution interacts with the subsurface both physically and
chemically. Laboratory analyses (rock-AMD reactivity, soil batch leach tests) and fieldwork (water
sampling, ground geophysics) were done to generate the input map layers and to validate the produced
The modified approach is illustrated through a discussion of a case study of AMD pollution in the
Witbank, Ermelo and Highveld coalfields of the Mpumalanga and KwaZulu-Natal Provinces in South
Africa. Many AMD cases have been reported in these provinces in recent years and are a cause of
concern for local municipalities, mining and environmental agencies. In the Witbank, Ermelo and
Highveld coalfields, several areas have been mined out while mining has not yet started in others,
hence the need to identify groundwater regions prone to AMD pollution in order to avoid further
impacts on the groundwater resources.
A knowledge-based fuzzy expert system, an artificial neural network (ANN) system and an adaptive
neuro-fuzzy inference system (ANFIS) were built using vulnerability factors (energy sources, ligands
sources, pollutant sources, transportation pathways and traps) to generate groundwater vulnerability
models of the coalfields. Highly vulnerable areas were identified in the Witbank coalfield and the
eastern part of the Ermelo coalfield which are characterised by the presence of AMD sources and good subsurface transport coupled with poor AMD pollution trapping properties. The results from all
three AI systems show a strong positive correlation of over 0.8 with groundwater sulphate
concentrations from two different datasets. In addition, relationships between geophysical responses
(resistivity and seismic velocity) and model values were obtained and used to validate the models.
The results show that the proposed approach can indeed be used for groundwater vulnerability
As a summary of the AI system, a toolbox containing all the tools needed for AI modelling of
groundwater vulnerability was developed. The toolbox is flexible and works both for manual and
semi-automatic processing allowing the user to contribute to the decisions taken in the modelling.
The toolbox serves to automate many processes associated with this type of modelling, resulting in
higher processing speed and improved accuracy in the generation of the model.
The methodology only considers the AMD pollution attenuation and migration at a regional scale and
does not account for local-scale sources of pollution and attenuation. Further research to refine the
approach may include the incorporation of groundwater flow direction and temporal datasets for the
future prediction of groundwater vulnerability. The approach may also be applied to other coalfields
to assess its robustness to changing hydrogeological conditions.||en_ZA
|dc.description.sponsorship||Department of Mineral Resources (DMR) of South Africa||en_ZA
|dc.publisher||University of the Free State||en_ZA
|dc.subject||Fuzzy expert system||en_ZA
|dc.subject||Artificial neural network||en_ZA
|dc.subject||Adaptive neuro-fuzzy inference system||en_ZA
|dc.subject||Thesis (Ph.D. (Institute for Groundwater Studies))--University of the Free State, 2017||en_ZA
|dc.title||Development of rapid assessment tools for groundwater vulnerability mapping using intergrated geoscientific datasets and artificial intelligence algorithms: case study from Witbank and Ermelo coalfields, South Africa||en_ZA
|dc.rights.holder||University of the Free State||en_ZA