Risk assessment is imperative to reduce consequences of accidents generated by leaks on pipelines. After a failure in an urban pipeline, the contents can be released rapidly forming a spreading pool in the surface or a vapor cloud that can lead to different events such as fires and explosions, affecting the population in the area. Generally, deterministic assessments are carried out where certain assumptions are made in the models used to calculate failure and event probabilities, event consequences and risk values. This causes the existence of uncertainty that is not considered and therefore it generates results only for “mean” or “worst-case” scenarios. Not taking uncertainty into account can lead to decisions with an unknown degree of conservatism: overestimate results could increase cost of decisions and underestimate results can lead to poor risk management. There is a lack of a clear directive to use methods of uncertainty assessment in risk analysis and for that reason, a framework is needed for carrying out quantitative risk assessment coupled with uncertainty treatment methods in order to handle, reduce, and properly inform uncertainty to the decision makers.
Uncertainty appears when we try to represent a system using different analysis techniques and therefore, depends on how accurately this process can be done. This in turn relies on how complex is the system under analysis and which techniques can be used. As part of the framework mentioned above, a classification is suggested to adequately select methods in probabilistic risk analysis to take into account different types of uncertainty in different kinds of representation and system analysis techniques. The classification is first based on the complexity of the system to be analyzed and then in the complexity of the modeling technique that is used to deal with system complexity. System complexity depends mainly on three variables: 1) understanding about the system (variables and its interactions) 2) the number of elements in the system 3) complexity of system interactions. Depending on system complexity, different techniques to represent it and analyze these systems in risk analysis are proposed and then, methods to propagate uncertainty are mentioned. A comparison of methodologies is presented to provide the engineer with tools to use adequately these methods in different situations and provide guidelines to deal with uncertainty in risk assessment.