Phillip Jordan, Alan Seed, Rory Nathan, Peter Hill, Eva Kordomenidi, Clive Pierce, Michael Leonard
This paper discusses the stochastic framework that was used to generate the 5449 sets of inflow hydrographs, to develop and stress test a dam operations model. The stochastic simulations were driven by 600 different space-time patterns of rainfall generated using a stochastic space-time multiplicative cascade model. Eight significant storms were identified in the radar archive to identify parameter sets for the stochastic generation algorithm and 600 replicates of space-time rainfall were generated. The statistical properties of spatial patterns of 48-hour rainfall bursts on eight major subcatchments of the Brisbane River catchment from the 600 stochastic replicates were verified against the same statistics derived from 38 major flood causing rainfall events observed in the catchment. The hydrographs were generated using an URBS rainfall runoff routing model of the Brisbane River catchment, which was calibrated to 38 historical flood events (between 1955 and 2013) and tested on a further 10 historical flood events (between 1887 and 1947).
The stochastically simulated sets of inflow hydrographs were then used to assess the impact of variations in flood operation rules for Wivenhoe and Somerset dams. The stochastically generated events exhibit substantial variability in runoff hydrographs but with variability that is statistically consistent with observed events. The stochastically generated hydrographs provide a considerably more realistic basis for testing the outcomes for different flood operations strategies than the single design event approaches that have previously been adopted.
Rory Nathan, Peter Hill
This paper provides an overview of the different simulation frameworks used for the estimation of design floods.. For small events the behaviour of many flood modifying factors is highly variable and chaotic, whereas as the magnitude of the event increases so does the organising influence of the dominant meteorologic conditions. The approach to design flood estimation will depend upon the availability of data and the exceedance probabilities of interest. The techniques can vary from frequency analysis of the data recorded at a site to rainfall-runoff modelling with design rainfall inputs derived from regional frequency analysis. For extreme floods, which are of relevance for assessing flood loadings for dams and the assessment of spillway adequacy, the stochastic (Monte Carlo) approach offers a number of advantages over the traditional deterministic approach. Although there has been significant progress in design flood estimation practice in Australia over the last couple of decades there remains many significant research and training needs.
Kristen Sih, Peter Hill, Susan Ryan, Siraj Perera
Although ANCOLD provides guidance on good dam safety practices, in Australia it is the State and Territory Governments’ role to protect the public from dam safety incidents and in many cases these jurisdictions have legally binding regulations in place that dam owners must adhere to. This paper presents a comparative analysis of the dam safety regulations currently in place for Australian states, as well as selected international jurisdictions. The limit of applicability of the regulations, number of dams regulated, content of the regulations and powers and responsibilities of the regulator are all compared. It was found that there is a large range within each of these categories with regulatory approaches varying from light-handed and objective based, to highly prescriptive. The extent to which risk management principles are used in the regulations for each jurisdiction has also been investigated. It was found that in jurisdictions where higher hazard category dams account for a higher proportion of dams being regulated, risk analysis is included in the regulations. Finally, the ANCOLD societal risk criteria and ALARP considerations have been compared and contrasted with those from international jurisdictions and other hazardous industries.
Nanda Nandakumar, Janice Green, Rory Nathan, Kristen Sih& Robert Wilson
A detailed assessment of hydrologic risk was undertaken for Hume Dam. Data available and relevant to the hydrologic risk assessment were collated and assessed. The catchment was divided into 35 different sub-catchments, each with its own set of parameters that characterised the local hydrologic response. Recorded streamflow was used to calibrate the flood response of selected gauged sub-catchments, and a combination of historic and synthetically-derived data was used to validate the model and loss parameters. The 35 models were combined into a single catchment-wide model. A Monte Carlo approach was adopted for the validation of the models and the derivation of Hume Dam inflow and outflow frequency curves. A range of PMFs which satisfy ANCOLD’s definition of the PMF were also estimated. The PMPDF outflow was estimated to be 7,600 m3/s which can be passed by the dam. Depending upon the assumptions made, the peak PMF outflow was estimated to be in the range from 10,300 m3/s to 14,900 m3/s
2011 – Assessment of Hydrologic Risk for Hume Dam
Chriselyn Kavanagh, Simon Lang, Andrew Northfield, Peter Hill
The U.S. Army Corps of Engineers have recently releasedHEC-LifeSim1.0, a dynamic simulation model for estimating life loss from severe flooding (Fields, 2016). In contrast to the empirical models that are often used to estimate life loss from dam failure, HEC-LifeSim explicitly models the warning and mobilisation of the population at risk, and predicts the spatial distribution of fatalities across the structures and transport networks expected to be inundated. This capability provides additional insights to dam owners that can be used to better understand and reduce the life safety risks posed by large dams. In this paper, we demonstrate the use of HEC-LifeSim to model the potential loss of life from failure of five large Australian dams. Particular attention is paid to how the predicted life loss varies with warning time, in a manner that depends on human response and the transport network’s capacity for mass evacuations, and the modelled severity of flooding. We also examine how the HEC-LifeSim estimates of life loss compare with those from the empirical Reclamation Consequence Estimating Methodology (RCEM).