Understanding, modeling, and reducing natural-hazard risks
|
Dec 2024, The University of Canterbury and University College London
Current seismic risk assessment tools fall short in capturing the complex interactions between people and the built environment in increasingly interconnected and urbanised societies. This limitation undermines the efficacy of such tools in supporting numerous critical risk-informed decision-making efforts, including but not limited to people-centred risk-management policy design. To overcome this shortcoming, bespoke models that capture the dynamic interdependencies between people and the built environment are required to be developed. Towards this aim, we mine large datasets of time-use surveys from different countries- which provide information on the various activities that people participate in at different times of the day - to develop data-driven models for predicting which people use various buildings and infrastructure for different services and/or activities, and when. The results of the data-driven models are then used to characterise an enriched seismic risk assessment approach, which leverages agent-based modelling to dynamically track interactions between people and the built environment as part of a nuanced, people-centred risk quantification process. The enriched risk assessment approach can produce a detailed representation of how individuals are affected by earthquake disasters, providing useful information for people-centred decision-making. The approach is demonstrated for Christchurch, New Zealand, considering a likely earthquake scenario nearby. This study contributes to advancing the state-of-the-art in seismic risk assessment tools and their decision-support capabilities. Collaborators: Tom Logan (Canterbury) and Gemma Cremen (UCL) |
|
May 2024, University College London and University of Edinburgh
Recent earthquake disasters have highlighted an urgent need for continuous advancements in approaches to reducing seismic risk. Decision-making on such strategies should consider subsurface geophysical information (e.g., seismic site response), given its direct link to seismic hazard. This is particularly important in regions where the poorest in society often reside in areas with softer soils that lead to higher ground-motion amplifications. In this context, we propose a framework to support decision-making on earthquake risk policies, which explicitly integrates information on the geophysics of an urban system as well as its physical and social environment. The framework is based on the Tomorrow’s Cities Decision Support Environment, which was designed to support urban planning with a focus on pro-poor disaster risk reduction in countries of the Global South. It is further underpinned by a cost-benefit analysis, which facilitates the assessment of potential policies in terms of both their ability to reduce earthquake risk as well as their value for (often limited) money. We illustrate the framework using a well-established virtual urban testbed based on Global South cities, which reveals that geophysics-informed policy making can successfully lead to pro-poor earthquake risk reduction. Collaborators: Himanshu Agrawal* (Edinburgh), Gemma Cremen (UCL), and John McCloskey (Edinburgh) *Shared first authorship |
|
March 2024, DE-RISC Lab, University College London
Disaster impact metrics (DIMs) are key outputs of natural-hazard risk models/assessments that provide a tangible way of communicating risk. However, typical DIMs are limited in that they tend to capture only direct damage/economic losses, be specifically designed for developed countries, account for just one snapshot in time, and be characterised for individual assets rather than systems. These shortcomings somewhat stem from a lack of understanding around the bespoke requirements of different stakeholders concerning disaster impact/risk assessments. Addressing these limitations, we propose a toolbox for characterising context-specific DIMs that capture relevant stakeholder priorities/requirements. The toolbox includes: (1) a comprehensive, holistic pool of DIMs developed from a literature review and a conceptual representation of how society functions; and (2) a stakeholder-centred framework for facilitating the appropriate selection of DIMs from this pool. We demonstrate the framework for Kathmandu, Nepal, revealing that the relative importance of a given disaster impact can change for different stakeholder groups and spatio-temporal dimensions. Impacts related to direct damage/economic losses are not the most crucial concern of the considered stakeholders. Higher priority is placed on characterising accessibility impacts around utilities and social networks, for instance. This work contributes to advancing the usefulness of natural-hazard risk assessments for important decision-making. Collaborators: Fabrizio Nocera, Gemma Cremen, Carmine Galasso (UCL), Vibek Manandhar, and Prayash Malla (NSET Nepal)
|
|
November 2023, DE-RISC Lab, University College London
Devastating earthquakes can cause affected households to relocate. Post-earthquake relocation disrupts impacted households’ social ties and, in some instances, their access to affordable services. Simulation-based approaches that model post-earthquake relocation decision-making can be valuable tools for supporting the development of related disaster risk reduction policies. Yet, existing versions of these models focus particularly on housing-related factors, which are not the sole driver of post-earthquake relocation. We integrate data-driven approaches and local perspectives to account for post-earthquake household relocation decision-making within an existing simulation-based framework for policy-related risk-sensitive decision support on future urban development. We use household survey data related to the 2015 Gorkha earthquakes in Nepal to develop a random forest model that estimates post-earthquake relocation inclination of disaster-affected households. The developed model holistically captures various context-specific factors important to the post-earthquake household relocation decision-making. We leverage the framework to quantitatively assess the effectiveness of various disaster risk reduction policies in reducing positive post-earthquake relocation inclination, with an explicit focus on low-income households. We demonstrate it using a future “Tomorrowville”, a hypothetical expanding urban extent that reflects important social and physical characteristics of Kathmandu, Nepal. Our analyses suggest that the provision of livelihood assistance funds is more successful when it comes to mitigating positive post-earthquake relocation inclination than hard policies focused on strengthening buildings (at least in the context of the examined case study). They also suggest viable pro-poor pathways for mitigating disaster impacts without the need to create potentially politically sensitive income-based restrictions on policy remits. |
|
January 2023, Tomorrow's Cities
Recent major earthquake disasters have highlighted the effectiveness of financial soft policies (e.g., earthquake insurance) in transferring seismic risk away from those directly impacted and complementing 'hard' disaster risk mitigation measures. However, the benefits of existing financial soft policies are often not guaranteed. This may be attributed to: (1) their low penetration rate (e.g., in the case of earthquake insurance); (2) the fact that they typically neglect the explicit needs of low-income sectors in modern societies, who are often disproportionately impacted by natural-hazard driven disasters; and/or (3) their failure to consider the time-dependent nature of urban exposure. We contribute towards addressing these shortcomings by proposing a flexible framework for designing and assessing bespoke, people-centered, household-level, compulsory financial soft policies (including conventional earthquake insurance, disaster relief fund schemes, income-based tax relief scheme, or a combination of those) across cities under rapid urban expansion. |
|
June 2022, Stanford Urban Resilience Initiative
Residential damage from major disasters often displaces local residents out of their homes and into temporary housing. Out-of-town contractors assisting in post-disaster housing reconstruction also need housing, creating additional pressure on the local housing stock. Communities should thus prepare for a surge in temporary housing demand to minimize the impact on the local residents and to expedite housing recovery efforts. This paper introduces an agent-based simulation framework to estimate the workforce demand and the joint temporary housing needs of contractors and displaced households. The framework can be used to evaluate the resulting challenges and benefits of interventions aimed at attracting out-of-town contractors to expedite housing recovery. We present a case study on the housing recovery of the city of San Francisco after hypothetical M6.5, M7.2, and M7.9 earthquakes. Collaborators: Rodrigo Costa and Jack Baker (Stanford University)
|
|
Feb 2022, Stanford Urban Resilience Initiative
This study investigates the effect of accounting for both physical damage to infrastructure and household disaster preparedness in estimates of potable water needs after earthquakes. A case study is presented involving the water supply system to the city of San Francisco after an M7.9 earthquake. Accounting for household preparedness helps identify regions in the city where water supply is interrupted, and many people may not have personal resources to access alternative sources of water. Considering both infrastructure disruption and household characteristics may inform decisions to allocate emergency water resources across the city during emergency response. Collaborators: Rodrigo Costa and Jack Baker (Stanford University)
|
|
|
March 2021, Course Project of CEE286 Stanford University
This study proposes a privacy-friendly, reliable, economical, and flexible smart home security monitoring system that consists of non-intrusive and low-cost vibration sensors, a stranger detection algorithm based on the two-sample t-test, and an occupant identification algorithm based on a support vector machine model. Unlike a traditional home security system that extensively deploys cameras, the proposed system uses the building structure to sense the rich information embodied in human footsteps indirectly and passes the information to vibration sensors. The proposed home security system is highly flexible in its definition of occupants and strangers, allowing the end-user to customize the system to satisfy various needs under different application scenarios. Previous studies have utilized footstep-induced floor vibrations for occupant detection and estimating the left-right walking gait of humans. However, studies focusing on stranger detection are scant. This study aims to address the urgent need for a privacy-friendly home security monitoring system using footstep-induced floor vibrations. Instructor: Haeyoung Noh (Stanford University)
|
|
November 2020, Course Project of CEE254 Stanford University
This study proposes an ensemble learning model with dynamic weighting scheme to predict the PM2.5 concentration at locations of interest. The model can perform short-term and long-term prediction, and interpolation tasks. The ensemble method is an amalgamation of six individual models of different levels of flexibility. Through bootstrapping, 100 samples are drawn from 28 days of data collected by both mobile and static sensors in Tianjin and 7 days of static and mobile data collected in Foshan. Six individual models are evaluated on the 100 samples to quantify uncertainties associated with each model. Weights of individual models consist of two parts: preliminary weights and dynamic weights. The preliminary weights are preassigned according to model performance on the 100 bootstrapped samples. The remaining weights are dynamically decided by either Cross-Validation error or error on the 20% held-out data during the learning process. The proposed ensemble learning model is expected to perform well on all three tasks for which it was designed. We won Best Project Award for Robust Prediction Model in the class competition. Collaborator: Ben Flanagan
Instructor: Haeyoung Noh (Stanford University) |
