Consideration of Network Vulnerability in Infrastructure Management Systems
A Erath, K W Axhausen, ETH Zurich, CH; R Hajdin, Infrastructure Management Consultants, CH; J Birdsall, EPF Lausanne, CH
This paper proposes a framework to implement vulnerability into infrastructure management systems. Applied for Swiss case,the possibilities of using subnetworks and statistical modelling to quantify additional user cost after failure is demonstrated.
Infrastructure disruption caused by natural hazards is a reality for transportation networks and might become even more important with an increase of such incidents due to climate change. However, only little research has been carried out so far to integrate vulnerability into the planning process and the management of transport infrastructures. Especially in developed countries the main attention of transport infrastructure development has shifted from network expansion and construction to the infrastructure management. To maintain the network efficiently infrastructure management systems (IMS) have been developed which determine optimal maintenance and improvement strategies for infrastructure components but they are not designed to integrate aspects of vulnerability. This paper presents a framework how vulnerability can be quantitatively integrated in the management of large road infrastructure networks and applies the methodology to the Swiss case. The project is carried out in two parallel streams. The first part identified hazard scenarios bringing together information about network components and natural hazard potential. The second part quantifies the additional user cost caused by a given hazard scenario and is the main focus of this paper.
For intensely used road networks in densely populated countries, like Switzerland, a reliable assessment of the additional user costs has to be based on traffic assignments which consider congestion effects. As the duration of most failures is assumed to be less than two weeks no effect on travel demand is considered. Mode choice effects are also neglected as those are assumed to be small over the network as a whole. A further constraint was that the hazardous network component assessment was carried out parallel to the stream reported here. Therefore, at the beginning of the project it was not clear which links are endangered. However, with the special topography of Switzerland it can be assumed that large parts of the network might be affected. So, it was decided that the methodology to be applied must be able to deliver estimates of the consequences of failure for all links in the national road transport model, which covers all main roads.
The national road transport model for Switzerland consists of around 20?000 links whose separate failures represent a case each. Given a calculation time of 50 minutes to reach equilibrium, full network assignments for all cases is unrealistic. Therefore an alternative approach was developed and validated. Analyses of full assignments showed that in most cases the effects of link failures are spatially restricted. The paper shows under what circumstances the application of subnetworks covering only a certain buffer around the affected link is appropriate, the accuracy that might be expected from such an approach and the optimal subnetwork size dependent on the local network topography and population and employment density respectively. Moreover, a statistical model is presented which enables authorities to estimate additional user costs caused by link failures without running an assignment model. This might be the case in developing countries where only information about the network, such as topology and traffic counts is readily available. The statistical model employs state-of-the-art methods accounting for spatial error and correlation effects and builds on earlier models predicting link speeds.
Association for European Transport