Dambreak & Consequences (September 2013) – MODULES 1 to 5
An understanding of the consequences of dam failure is essential in dam safety emergency planning and as an input to risk assessment. In recent years there has been significant advances in hydraulic modelling and access to high quality elevation data which has revolutionised dambreak modelling. The advent of risk based approaches has increased the focus on estimating the consequence of dam failure and particularly the potential loss of life. The method developed by the USBR in 1999 has had widespread application in Australia and in recent years a number of more sophisticated simulation approaches have been developed. This session will cover the latest developments in dambreak modelling and the estimation of potential loss of life from dam failure.
This course is designed to present the state of practice on these matters for dam safety risk management. The 2 days are designed for both experienced and less experienced dam owners, regulators and consultants.
Includes access to the following videos:
$0.00 - $80.00
The ANCOLD (2003) Guidelines on Risk Assessment contain criteria regarding the tolerable level of individual risk from dam failure. Maslin et al. (2012) describe an approach to estimating individual risk from dam failure, using exposure factors, warning and evacuation factors, and fatality factors. These factors vary according to the people at risk, the anticipated warning time, the flood severity and the shelter people are likely to be in. Maslin et al. (2012) provide step-by-step instructions, which means their approach can be applied in a consistent manner from dam to dam. However, the recommended fatality factors are based on Graham (1999) and DHS (2011) definitions of high, medium and low severity flooding which have been superseded by the Reclamation Consequence Estimating Methodology (RCEM). Therefore, in this paper modifications to the Maslin et al. (2012) approach are proposed, so that estimates of individual risk from dam failure are consistent with RCEM-based estimates of societal risk. The paper then concludes with two predictions about how the assessment and use of individual risk in Australian dam safety management may change in future.
Simon Lang, Peter Hill, Wayne Graham
The empirical method developed by Graham (1999) is the most widely used in Australia to estimate potential loss of life from dam failure. It is likely to remain that way while spatially based dynamic simulation models are not publicly available (e.g. LIFESim, HEC-FIA and LSM). When the Graham (1999) approach was first developed the prevalence of spatial data and the speed of computers was much less. In addition, most people did not have mobile phones, social media was in its infancy, and automatic emergency alert telephone systems were 10 years from being used in Australia. Graham (1999) was intended to be applied to populations at risk (PAR) lumped into a discrete number of reaches. The selection of fatality rates for the PAR in each reach was based on average flood severity and dam failure warning times. Today, there is typically much more spatially distributed data available to those doing dam failure consequence assessments. Often a property database is available that identifies the location of each individual building where PAR may be, along with estimates of flood depths and velocities at those buildings. News of severe flooding is likely to be circulated by Facebook, Twitter and e-mail, in conjunction with official warnings provided by emergency agencies through radio and television and emergency alert telephone systems.
This raises the question of how Graham (1999) is best applied in today’s digital age. This paper explores some of the issues, including the estimation of dam failure warning time, using Graham (1999) to estimate loss of life in individual buildings and the suitability of Graham (1999) for estimating loss of life for very large PAR.
Keywords: loss of life, dam safety, risk analysis.
David Stephens, Peter Hill, Rory Nathan
The estimation of incremental consequences of dam failure often requires consideration of coincident flows in downstream tributaries. In the past overly simplistic assumptions have often been adopted. Examples include an assumption that flows in downstream tributaries are negligible, equivalent to the 1 in 100 Annual Exceedance Probability (AEP) flood, the mean annual flood or the flood of record. Experience has shown that these assumptions often underestimate coincident flows, particularly for extreme events approaching the AEP of the Probable Maximum Precipitation. Additionally, the justification for adopting these techniques is usually driven by ease of use rather than the degree to which they represent the relevant physical processes at play. For some dams, these techniques may have a negligible influence on the overall consequence assessment. However, there are many dams for which an improved understanding of coincident flows using a joint probabilistic framework can result in significantly altered estimates of the natural flood and dambreak flood inundation zone. This can frequently lead to the consequences of the natural flood being larger than would otherwise have been the case, leading to a reduction in incremental consequences. Two examples of such situations are presented, including a description of the techniques used to estimate coincident flows and a discussion on likely influence of these flow estimates on incremental consequences. These examples are then used to draw some general principles for the types of dams at which an improved understanding of coincident flows is warranted.
Keywords: dam failure, coincident, joint probability, consequence assessment
Simon Lang, Chriselyn Meneses, Peter Hill, Kristen Sih
In Australia to date, the empirical method developed by Graham (1999) is the most widely applied approach for estimating loss of life from dambreak flooding. However, as the move to risk-based approaches of dam safety management has gathered momentum internationally, increasingly sophisticated techniques for estimating loss of life have emerged. One of these models is the United States Army Corps of Engineers (USACE) HEC-FIA model. HEC-FIA models the influence of flooding, structure characteristics, and warning and evacuation assumptions on loss of life in a spatially distributed manner. In contrast to Graham (1999), HEC-FIA also allows the user to model the loss of life for both dambreak and natural flooding.
This paper presents the results from the first Australian application of HEC-FIA to two dams in southeast Australia. The application of empirical methods developed by Graham (2004) and Reiter (2001) is also discussed.
Simon Lang, David Stephens, Peter Hill, Mark Arnold and Tommie Conway
Considerable thought has been given in recent years to managing the risks associated with floods during the construction of new dams and dam upgrades. Both ANCOLD and the NSW DSC provide some limited advice on how this risk should be managed, with many dam owners aiming for societal risk during construction to be no higher than pre-construction. One approach to do this is to draw down the reservoir such that sufficient airspace is created to reduce the probability of overtopping the construction works to be equal to that of overtopping the dam crest pre-construction. However, this frequently leads to very large releases of valuable water resource being required. This approach also fails to consider that the conditional probabilities of failure may be quite different during construction than during normal operation. A risk-based approach was applied for the recent upgrade of Tarago Reservoir. Existing event trees from a failure modes analysis were adjusted to reflect the construction conditions. In some cases, the event probabilities increased (for example as a result of excavation of the dam embankment), however some also decreased (for example as a result of more rapid means of detecting and intervening in breach formation during construction). The conditional probabilities of failure during construction were then used to estimate the overall seasonal probability of failure, and it was found that a limited draw down of the reservoir would be sufficient to ensure that risks were no higher during construction than pre-construction. To reinforce this, the cost-to-save-a-statistical life was estimated for further drawdown of the reservoir and used to demonstrate that the risks were as low as reasonably practicable.