Water infrastructure vulnerability due to dependency on third party infrastructure sectors

Abstract

Society is critically dependent on water networks to provide a safe and reliable supply of clean water. Water companies must balance the requirement for very high levels of reliability with the need to protect the customer from excessive bills. This, in turn, requires them to assess the entire range of risks including low probability, high impact events. Water companies are confident when estimating the risk of failure due to natural hazards affecting their own systems but they are less certain assessing how failures in third party infrastructure could affect their services. For example, 1 500 homes in Cumbria lost their water supply in 2005 due to power cuts caused by flooding and strong winds. This highlights the need for better methods to help water companies assess these risks and their options for managing them. Current risk assessments rely heavily on the expertise, experience and intuition of companies’ employees. However, the interactions between different infrastructure networks create complex systems which can behave unpredictably and leave customers vulnerable to unanticipated consequences. Previous academic studies have been hampered by limited data and therefore have mainly used coarse resolution models which simulate only the high-level performance of idealized networks. This thesis has improved on this situation by developing more realistic models of water systems and their dependencies. Two real-world case studies have been used to explore their potential as aids to inform better decision making. The first model assesses the likelihood and consequence of third party infrastructure failures causing water supply interruptions. It draws on catastrophe modelling techniques used in the insurance industry and is composed of three elements: i) a hazard model producing synthetic but realistic time series of wind, temperature and rainfall; ii) a suite of fragility curves describing the susceptibility of highways, electricity, telecommunications and water facilities to these hazards; and iii) a set of network models to explore the impact of individual facility failures on the availability of water supplies. ii The model is implemented for a real-water distribution network where dependence on external infrastructure systems was found to cause an expected loss of 9.9 minutes per property per year. In isolation, electricity, telecommunications and transport respectively make up 75%, 11% and 0% of the total risk. The remaining 14% results from interactions between these sectors. It is argued that these failure modes are unlikely to be identified using existing risk assessment methods. While the first model provides a quantitative and probabilistic risk assessment, its complicated nature makes interpreting the results challenging and limits the number of scenarios that can be investigated. The second model takes a different approach and focuses upon identifying low probability, high impact events. The hazard model and fragility curves are replaced by the UK Cabinet Office’s ‘reasonable worst case scenarios’, and a simpler stocks and flow model takes the place of the hydraulic network models but maintains the representation of the network structure and components. The results provide insight into how, where and why water supplies are vulnerable to failures in third party sectors. They show that dependencies can dramatically increase vulnerability (in one case the loss of power to an emergency pumping station causes 6 million property hours without supply). Equally, an inland flooding scenario shows that simple solutions such as installing a connection for a mobile generator can significantly reduce vulnerabilities. The methods developed in this research make a significant contribution to closing the gap between existing theoretical studies of dependency and the requirements of infrastructure providers to improve the resilience of real systems. The first model provides a probabilistic assessment of risk that enables infrastructure providers to prioritise investment. The second model identifies the full range of vulnerabilities and investigates the sensitivity of the outputs to model parameters and inputs.