Bi-phase methodology for sensitivity analysis of complex models, applied to the model of evaluating resilience in transport networks

Sensitivity analyses identify the importance of the model parameters and variables, providing a profound knowledge of the model. Most of the sensitivity analysis implies a local evaluation of the variability of the parameters.... [ view full abstract ]

Sensitivity analyses identify the importance of the model parameters and variables, providing a profound knowledge of the model. Most of the sensitivity analysis implies a local evaluation of the variability of the parameters. However, in those cases where the model involves a large number of degrees of freedom, these studies become highly time consuming and incapable of obtaining conclusive solutions. This paper presents a bi-phase approach, by integrating a local into a global sensitivity method. This methodology is recommended in those multidimensional models that make other approaches inefficient. The global phase is based on the statistical distribution of the variables, providing a robust and reliable solution, with low computational costs. With the aim of showing the potential of this approach, the method is applied to a complex existing model. This model aims to evaluate the resilience in transport networks when affected by an extreme weather event. As expected, this sensitivity analysis shows the large influence of the hazard intensity, but also shows how other variables, as the size of the area affected by the hazard and the role that the users play, can modify the values of the sensitivity analysis significantly.