A Probabilistic Methodology for DBA and its Comparison to Current Methods

Abstract

Dam break assessment (DBA) is a widely used method to estimate the potential consequences of an eventual failure of a tailings dam on the downstream affected area, including inundation extent, characteristics, and flood arrival times. Release volume and peak flow at moment of failure are key input parameters of the assessment, as the magnitude of the downstream consequences is directly proportional to them. However, variability in tailings rheology and breach characteristics highly influence these two inputs, their variability acting as a source of uncertainty for the estimation of their values. 

In order to account for this uncertainty, a probabilistic methodology has been used in this paper to perform a case-study DBA in which key variables affecting failure characteristics are treated as random variables that follow a given probabilistic distribution. A statistical analysis of past tailings dam breaches obtained from a public database was carried out in order to define the distribution assigned to each variable. Additionally, an assessment of available case-specific tailings properties that play a role in release volume and peak flow was also performed to estimate their variability. 

Through a predefined and parameterized failure surface, a Montecarlo simulation process was applied to obtain the values of release volume and peak flow with associated exceedance probabilities. DBA models were built and run from relevant scenarios arising from the Montecarlo simulation, whose results were compared to the ones obtained by the deterministic methods most widely used nowadays. The methodology was validated by adopting a past well-known breach as the case-study for this paper.

Authors: 

Francisco Moyano | SRK Consulting, Salta, Argentina 

Federico Giurich | SRK Consulting, Salta, Argentina

Presenter: 

Francisco Moyano | SRK Consulting, Salta, Argentina