A raw water security risk model for urban supply based on failure mode analysis

•A new model of water security risk is proposed for urban supply.•Security indicators are coupled with risk variables to quantify the risk.•The model is tested in a large metropolitan region (Belo Horizonte, Brazil).•Model results expose severe and frequent water demand, drought and contamination risks.•Results provide insights to prioritize adequate mitigation measures. The provision of raw water to urban supply systems is a global endeavor, and a great challenge to water resource managers of large metropolitan regions. A sustainable supply requires the integrated management of water quantity and quality, as well as the system’s adaption to land use and climate changes (e.g., deforestation, droughts). The available water security models inform about the current security status of a system based on the assessment of quantity, quality and adaption indicators. But they barely include risk variables in the analysis, which could help to improve the security assessment considering the historical records of indicators or their future projections. In the present study, a new method is developed that couples a security assessment based on the rating of pressure indicators acting on a watershed, such as water demand, ordinary and accidental contaminants, droughts, and environmental settings (e.g., share of forest cover), with a risk assessment based on pressure properties such as severity, occurrence and detectability. The method is framed in the “Failure Mode and Effects Analysis”, and was applied to the Rio das Velhas system that supplies a portion of Belo Horizonte metropolitan region (Minas Gerais, Brazil) with >4 million inhabitants. The results exposed unacceptable risks for water demand, ordinary contaminants and droughts, because of their high severity and frequency. The water demand risks were explained by data on current (2019) water scarcity and population growth projections (until 2036) that are likely to raise water consumption. The ordinary contaminant risks were deduced from information on current deviations of water quality parameters (e.g., dissolved oxygen, turbidity) from legal thresholds, and records (since 2007) of stream water contamination by untreated domestic effluents. Finally, the drought risks were explained by information on the persistence of stream flows below a legal reference flow, coupled with the identification of many (23) dry years in the available stream flow record (48 years). Overall, the proposed model was efficient in the e