Building an annotated damage image database is the first step to support AI-assisted hurricane impact analysis. Up to now, annotated datasets for model training are insufficient at a local level despite abundant raw data that have been collected for decades. This paper provides a systematic approach for establishing an annotated hurricane-damaged building image database to support AI-assisted damage assessment and analysis. Optimal rectilinear images were generated from panoramic images collected from Hurricane Harvey, Texas 2017. Then, deep learning models, including Amazon Web Service (AWS) Rekognition and Mask R-CNN (Region Based Convolutional Neural Networks), were retrained on the data to develop a pipeline for building detection and structural component extraction. A web-based dashboard was developed for building data management and processed image visualization along with detected structural components and their damage ratings. The proposed AI-assisted labeling tool and trained models can intelligently and rapidly assist potential users such as hazard researchers, practitioners, and government agencies on natural disaster damage management.
more »
« less
A New Data-Driven Approach to Measuring Hurricane Risk
Improving disaster operations requires understanding and managing risk. This paper proposes a new data-driven approach for measuring the risk associated with a natural hazard, in support of developing more effective approaches for managing disaster operations. The paper focuses, in particular, on the issue of defining the inherent severity of a hazard event, independent of its impacts on human society, and concentrates on hurricanes as a specific type of natural hazard. After proposing a preliminary severity measure in the context of a hurricane, the paper discusses the issues associated with collecting empirical data to support its implementation. The approach is then illustrated by comparing the relative risk associated with two different locations in the state of North Carolina subject to the impacts of Hurricane Florence in 2018.
more »
« less
- Award ID(s):
- 1735139
- PAR ID:
- 10110723
- Date Published:
- Journal Name:
- Proceedings of the 16th ISCRAM Conference
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
more » « less
Abstract Conceptualizing, assessing, and managing disaster risks involve collecting and synthesizing pluralistic information—from natural, built, and human systems—to characterize disaster impacts and guide policy on effective resilience investments. Disaster research and practice, therefore, are highly complex and inherently interdisciplinary endeavors. Characterizing the uncertainties involved in interdisciplinary disaster research is imperative, since misrepresenting uncertainty can lead to myopic decisions and suboptimal societal outcomes. Efficacious disaster mitigation should, therefore, explicitly address the uncertainties associated with all stages of hazard modeling, preparation, and response. However, uncertainty assessment and communication in the context of interdisciplinary disaster research remain understudied. In this “Perspective” article, we argue that in harnessing interdisciplinary methods and diverse data types in disaster research, careful deliberations on assessing
Type III andType IV errors are imperative. Additionally, we discuss the pathologies in frequentist approaches, calling for an increasing role for Bayesian methods in uncertainty estimations. Moreover, we discuss the potential tradeoffs associated with information and uncertainty, calling for deliberate consideration of the role of diversity of information prior to setting the scope in interdisciplinary modeling. Future research guided by further reflections on the ideas raised in this article could help push the frontiers of uncertainty estimation in interdisciplinary hazard research and practice. -
Adrot, A. ; Grace, R. ; Moore, K. ; Zobel, C. W. (Ed.)The devastating economic and societal impacts of COVID-19 can be substantially compounded by other secondary events that increase individuals’ exposure through mass gatherings such as protests or sheltering due to a natural disaster. Based on the Crichton’s Risk Triangle model, this paper proposes a Markov Chain Monte Carlo (MCMC) simulation framework to estimate the impact of mass gatherings on COVID-19 infections by adjusting levels of exposure and vulnerability. To this end, a case study of New York City is considered, at which the impact of mass gathering at public shelters due to a hypothetical hurricane will be studied. The simulation results will be discussed in the context of determining effective policies for reducing the impact of multi-hazard generalizability of our approach to other secondary events that can cause mass gatherings during a pandemic will also be discussed.more » « less
-
Pre-positioning of supplies is important to facilitate disaster relief operations, however it is only after a disaster event occurs that the effectiveness of the pre-positioning strategy can be properly assessed. With this in mind, this paper analyzes a risk-based pre-positioning algorithm, developed for the American Red Cross, in the context of its actual performance in the 2013 Colorado Front Range floods. The paper assesses the relative effectiveness of the pre-positioning approach with respect to historical asset placements, and it discusses changes to the model that are necessary to support such comparisons and allow for further model extensions.more » « less
-
more » « less
Abstract We develop a computational framework for the stochastic and dynamic modeling of regional natural catastrophe losses with an insurance industry to support government decision‐making for hurricane risk management. The analysis captures the temporal changes in the building inventory due to the acquisition (buyouts) of high‐risk properties and the vulnerability of the building stock due to retrofit mitigation decisions. The system is comprised of a set of interacting models to (1) simulate hazard events; (2) estimate regional hurricane‐induced losses from each hazard event based on an evolving building inventory; (3) capture acquisition offer acceptance, retrofit implementation, and insurance purchase behaviors of homeowners; and (4) represent an insurance market sensitive to demand with strategically interrelated primary insurers. This framework is linked to a simulation‐optimization model to optimize decision‐making by a government entity whose objective is to minimize region‐wide hurricane losses. We examine the effect of different policies on homeowner mitigation, insurance take‐up rate, insurer profit, and solvency in a case study using data for eastern North Carolina. Our findings indicate that an approach that coordinates insurance, retrofits, and acquisition of high‐risk properties effectively reduces total (uninsured and insured) losses.
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
