Abstract:
Surface-strip coal mine rehabilitation planning in South Africa and Australia is immature.
Rehabilitation risk assessment, despite being advocated by leading practice guidelines
and in some instances by legislation, is conducted with minimum requirements often met
by rehabilitation professionals. Specialist data is gathered during mine approval and for
the environmental impact assessment process. However, the focus of this is toward
assessing mining impacts and not for rehabilitation risk assessment. Quantitative,
integrated, multi-disciplinary rehabilitation risk assessment is seldom undertaken. This
thesis provides a methodology towards the development of a quantitative, integrative,
multi-disciplinary rehabilitation risk assessment model. Its purpose being to 'profile'
surface-strip coal mine sites, in terms of their rehabilitation risk and potential for
rehabilitation failure, from the outset of mine operations, with adjustments possible
progressively during mine operations. The methodology was developed by first
reviewing techniques suitable for the development of the model, as well as techniques
developed by others. Bayesian networks (BN) were found to be the most suited. A R2AIN
framework was then provided as a process towards developing several BN risk event
models that can amalgamate to form a synthesis rehabilitation risk assessment model. A
case study soil compaction BN model was used to demonstrate the framework in South
Africa and Australia. The case study showed that it is possible to integrate and quantify
rehabilitation risk, and most importantly to segregate risk into discrete contributing multidisciplines
for analysis. Risk percentages can be calculated per multi-discipline, per mine
phase, per site, to aid site risk ‘profiling’. It is recommended that further risk event BN
models be prioritised for development and that a rehabilitation risk assessment model be
developed to synthesise these into one model. This will require continuous improvements
in the method, to build confidence, including extensive risk event and synthesis BN
model evaluation and testing; improved BN input node states and values; and
simplification of the conditional probability table construction method. Adaptation to
other mining types, development activities and other regions should be investigated, as
well as spatial linkages to geographic information systems. This research contribution
improves upfront mine rehabilitation planning and decision making, providing improved
tools and techniques than what currently exist.
