Risk screening for dam break failures

Abstract

Dam failures are rare but highly catastrophic events which can be devastating for both infrastructure and the population living downstream. The recent failures of Abu Mansour dam in Libya (September 2023) and Kakhovka dam in Ukraine (June 2023) have shown that large number of fatalities are associated to these events. It is therefore important to understand the hydrodynamic behaviour of dam break waves to ultimately identify the risk they pose, especially for dams which do not have detailed data information for modelling studies. The objective of the present study is to determine the hydraulic characteristics of the waves and to predict the potential loss of life in case of a dam failure event using a simplified screening method which can be applied to a larger dam dataset. The approach of the research was to utilise the original theory of Ritter (1892) who presented an idealised solution to the Saint-Venant equations. These equations were modified by incorporating the effect of friction which eventually had a reducing effect on the idealized dam break wave celerity. Using this simplified hydraulic model the dam break wave depths and velocities were able to be determined from the required input of dam height. From additional past dam failures, the proposed limits of fatality rates in the curves of Reclamation Consequence Estimation Methodology (RCEM 2014) were validated. To determine the Population at Risk (PAR) the study utilised Geographic Information System (GIS) tools and buffer methods were tested. For a smaller dataset of 22 dams, selected based on range of dam heights from the Global Reservoir and Dam (GRanD) database, the results showed that a buffer with downstream distances determined by one hour travel time and a 1000 m width gave the most consistent estimates of the exposed population when compared to the method incorporating Digital Elevation Models (DEM). When the analysis was extended to the complete dataset of GRanD, a new buffer method was proposed because of the limitation of GIS analysis tool. This buffer method with downstream distance of 25 km was tested against three different inundation widths. Results showed that using a buffer with fixed downstream distance of 25 km and an inundation width of 1000 m gave the most closest estimates when compared with the DEM method. The study was able to highlight the total population at risk on a global scale and also zoomed into the region of Europe. Comparison between the dams allowed identification of high risk dams in terms of the exposed population. Specifically, the total potential fatalities were determined for the top ten PAR dams of Europe. Furthermore, to demonstrate the usability of the risk screening method the analysis was extended to two developing countries, Pakistan and Madagascar, where Multi Hazard Early Warning Systems (MHEWS) do not tend to exist. The results showed the difference in PAR and fatalities for each dam and provides valuable insights to develop preliminary risk management and evacuation plans both at local and national level. Despite its limitations, the study provides a quick risk screening method which can serve as an alternate to the resource intensive numerical modelling approach. Future research might be needed to reduce uncertainties and achieve more accurate estimates.