NICE (Networked Infrastructures under Compound Extremes)

Research Members: Abdullah Alhadlaq, Aparimit Kasliwal, Ariel Salgado

NICE: Networked Infrastructures under Compound Extremes

As climate change continues and technology advances, facilities around the world face ever increasing threats from natural hazards, cyber attacks, etc. The interconnectivity of systems has the potential to exacerbate resultant system failures. When one system fails, any system which depends on it also fails, producing a cascade effect from one network to the next. Such events have been observed in the power grid and other interdependent infrastructure networks amplifying blackouts. Furthermore, the possibility of compound events, like malicious actors launching a cyber-attack during a natural disaster, magnify the risk to already stressed networks. In order to mitigate this risk, we are developing new computationally tractable theoretical frameworks and methods to help ensure installation-level resilience. We are implementing Multiplex Network Science (MNS) to capture the structural properties of the internal network and Multiscale System Dynamics (MSD) to investigate the infrastructure response to compound extremes. In particular, we are focusing on mapping failure and recovery pathways, adapting to changing conditions, and recovering from disruptions.  Upon completion, we aim to have developed an integrated resilience framework informed by a suite of models, embedded in a proof-of-concept software. This will be capable of continuously monitoring the state-of-resilience of a facility and understanding failure and recovery pathways at multiple scales in order to minimize risk and downtime.

Multi-Scale Traffic Congestion Spreading Through Contagion Models

This work focuses on applying Multiscale System Dynamics (MSD) to traffic congestion by developing a theoretical framework that utilizes multi-scale dimension reduction to analyze the spread of traffic congestion. Recent efforts have centered on generating realistic traffic congestion patterns in the Bay Area, using infection spreading models calibrated at different geographical scales such as census tracts and counties to reproduce these patterns. Estimates at various scales are linked to provide an overall consistent assessment, enabling the multiscale modeling framework to simplify system dynamics by considering scale and offering indicators to prioritize actions in different geographic areas. The methodology involves constructing spatial regions of analysis at different spatial resolutions using Uber H3, allowing for more granular regions of analysis beyond just census tracts and counties. Previous studies have demonstrated the effectiveness of the SIR (Susceptible-Infected-Recovered) epidemic model in explaining traffic congestion spread, and this study applies the SIR model spatially to prioritize actions at different scales. The primary challenge is the noise in dynamics at finer scales, which the team aims to address by using dimension-reduction techniques to connect macro and micro-level dynamics. Daily traffic simulations measure resulting congestion, with the observed congestion data used as a proxy to calibrate the spatially distributed SIR model.