The issue of subsidence arising from underground coal mining resources has been accompanied in recent years by a number of unexpected and publicised subsidence incidents in NSW.

One explanation for the occurrences of the incidents is that subsidence is an “inexact science” due primarily to the inherent uncertainties in the nature of subsidence-related risks and the effectiveness of risk control measures.

Uncertainty in subsidence management is inherent. The complexity of subsidence engineering, that involves information from multiple disciplines and from multiple phases of investigations, can make it difficult to make the right risk management decisions.

Statistical and/or human error are other contributors to this uncertainty as is a biased approach to subsidence engineering or management. This latter point may be demonstrated by a tendency towards statistical analyses of surface subsidence movements without adequate consideration of the mechanisms of ground deformations.

Without a systematic assessment of uncertainties as a key factor in risk-based subsidence management, inappropriate decisions may cause valuable resources to be wasted, while significant hazards may go unrecognised.

A basic risk management process consists of:

i) Determining the required or desired management outcomes;

ii) Monitoring the actual system performance;

iii) Comparing the system performance with the required or desired management outcomes, and

iv) Taking actions to correct deviations of the system’s performance from the required or desired outcomes.

Uncertainties should be ranked according to:

i) Consequences of subsidence impacts as a result of the deviations, and

ii) The nature of the deviations. In particular, are such deviations reducible (or manageable)?

Uncertainties which may lead to irreducible deviations and severe consequences should warrant the highest priority for risk treatment. For example, undermining an important infrastructure located above the installation face of a longwall panel may represent such a significant challenge. In this case, once the surface movements become observable after the longwall face has passed the “square” position, any deviations, such as subsidence exceeding any defined tolerances, could become unstoppable.

It is possible to distinguish between two types of uncertainties for engineering purposes.

Type A uncertainties would be due to random variations caused by numerous factors such as natural variations in site conditions, sample sizes or measurement errors, etc.

Type B uncertainties come from imperfect knowledge. These may include hazardous conditions that are unrecognised, unknown or unsuspected, due to various reasons such as poorly planned investigations or lack of understanding/experience, etc. The resulting subsidence impacts are “unplanned” and thus unprepared for. For instance, the impacts of subsidence on the Cataract River and the associated upsidence in the river valley occurred some years ago as a “surprise”

Another may be known hazardous conditions that are, however, unpredictable or unquantifiable within the limits of practical constraints. A third is known hazardous conditions that are incorrectly or imperfectly understood, which may lead to inaccurate characterisation of risks.

Type A uncertainties (ie. random variations) will be dominant in the lower consequence regime and the conventional risk assessment approach (probability times consequence) is likely to be adequate for risk management purposes.

Type B uncertainties on the other hand generate multiple risk scenarios. This occurs once the data required for assessment becomes vague and scarce, as it invariably does as the consequences become higher.

A common approach to the management of uncertainties in the high consequence regime is to err on the side of caution. However, without understanding the nature of uncertainties and especially the resulting risk scenarios, implementation of a “conservative bias” may not necessarily be effective.

Management of subsidence impacts with potentially high consequences, where Type B uncertainties may become dominant, is a major challenge for the future.

The key is to develop a flexible and responsive management system that entails:

i) Progressive improvement of understanding through specifically designed monitoring and on-going reviews of the system’s performance and its major assumptions (apart from research and systematic database development, etc.);

ii) Progressive re-engineering of the management system to respond to changes or up-dated information/understanding. The re-engineering may involve planned mitigation (including layout options) and remediation measures, and

iii) A management plan with a monitoring program capable of detecting early warnings and an adequate contingency that will be effective under various identified risk scenarios.

Clearly, it will not be easy to fully implement all the above measures, due to difficulties associated with inflexibility of certain mining systems, such as a longwall, or the need for risk control measures which can be very costly and significant in design/planning terms.

There have been a number of cases involving subsidence impacts with potentially high consequences that comprised the following elements:

i) Subsidence predictions indicating acceptable risks, according to reviews jointly conducted with relevant stakeholders;

ii) Simulation of subsidence development (ie. pro-active monitoring) at sites with similar ground characteristics, but away from the actual target surface features to be affected by subsidence;

iii) Review of the planned subsidence management plan according to the results of subsidence predictions and site simulations, and

iv) Implementation of the reviewed subsidence management plan at the actual site to allow mining to take place under or near the target surface feature.

In conclusion, uncertainties in subsidence management have not been subject to systematic studies, despite the fact that they are inherent in almost every step of subsidence management processes.

Based on the nature of uncertainties encountered frequently in subsidence management, there appear to be two basic types of uncertainties, one of which is not the subject of a probabilistic approach.

Consequently, conventional risk assessment may not be adequate for this type of uncertainty if the potential consequences are likely to be high. In this case, we should acknowledge the validity and importance of the plausible risk scenarios and address them with a flexible and responsive risk management process.

* Article based on a paper presented at the Subsidence Management conference in October 2004. For additional details contact gang.li@minerals.nsw.gov.au