Skip to main content Skip to secondary navigation

A map of San Francisco with highway bridges marked as colored dots. The deeper the red of a bridge’s dot, the more influential it is with respect to the particular performance metric considered in the analysis.

Prioritizing Bridge Retrofits in Regional Road Networks under Seismic Hazard

Bridges are an integral and especially vulnerable part of regional road networks in seismically active areas. A key concern of managing such road networks is identifying which bridges to retrofit given a limited budget. Retrofitting bridges reduces the likelihood and consequences of disruptions in the operation of the road network, which can result in billions of dollars in losses.

In large networks, evaluating each possible combination of bridge retrofits is computationally intractable. In this work, we develop a heuristic method for bridge retrofit selection based on global variance-based sensitivity analysis. Building on the framework of probabilistic seismic risk assessment, this method links the performance of the road network to the states of its constituent bridges. The performance of the road network is a user-defined quantity, e.g. the expected indirect cost, median welfare loss, or expected decrease in accessibility. The analysis accounts for network effects, i.e., the effects of retrofitting particular combinations of bridges that are not simply the sum of the effects of retrofitting individual bridges.

The method produces a ranking of bridges for retrofit prioritization that is probabilistic, that preserves the heterogeneity of the network components, that is computationally tractable for complex regional networks, and that links the performance of the road network to the states of its components.

Though developed with respect to bridges in road networks under seismic hazard, this work should be extensible to other types of complex networks subject to different hazards. In broader terms, this work supports resilience-minded managers of large, spatially-distributed infrastructure networks that are subject to uncertain hazard in making component-level decisions with respect to the probabilistic performance of the network as a whole.


Gitanjali Bhattacharjee

Jack Baker