RecovUS: An agent-based model of post-disaster housing recovery

Date

2020-05

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

Population growth in hazard-prone areas coupled with increase in severity of extreme events have raised the potential for disaster-induced losses. This necessitates a better understanding of recovery process. Extensive analysis capabilities and modeling is required to capture dynamics of recovery and to underpin recovery policies. Among different aspects of recovery, housing restoration is of vital importance as it has a ripple effect on the overall recovery of a community. Housing is a primary element of peoples’ lives and influences their well-being. Residential structures constitute the major share of building stock in the United States. Additionally, neighborhood characteristics are influenced by households’ preferences and demands. A multitude of parameters influence housing recovery. These parameters can be classified as internal, interactive, and external drivers of recovery. Several models have been proposed that use a number of these drivers to simulate the recovery behavior. Households’ recovery decisions, however, are impacted by many parameters that their complete inclusion in a single model is infeasible. There are still gaps in the full understanding of post-disaster recovery and of how decisions made by individuals and different entities interact to output the overall recovery of a community. Additionally, integrating spatial aspects of recovery into models is an essential key which is missing in many studies. The current research aims to bridge the gap by developing a spatial model for simulation and prediction of homeowners’ recovery decisions through incorporating recovery drivers that could capture interactions of individual, communal, and organizational decisions. RecovUS is a spatial agent-based model for which all the input data can be obtained from publicly available and free sources. This research targets households living in their owned primary single-family detached houses, the prevalent housing type in the United States. In this study, a household’s recovery decision can be repair/reconstruction of its damaged/destroyed home, waiting without repair/reconstruction, or selling the home (and relocating). The research assumes that housing recovery is a function of households’ financial conditions and their community’s recovery conditions. The developed model, RecovUS, is founded on the assumption that a household would have the chance of repair/reconstruction if 1) it has enough financial resources, and 2) its community has recovered adequately. RecovUS models the influence of financial aids received from disaster insurance policies and public agencies. It also simulates the effect of recovery of community, including infrastructure, neighbors, and community assets, on households’ recovery decisions. RecovUS is a spatial agent-based model in the sense that it includes spatial locations of homes and community assets and captures interactions of households with their neighbors and perceived community assets. Based on these assumptions and interactions, the model predicts if a disaster-affected household decides to repair/reconstruct, wait, or sell its residence and if a buyer decides to repair/reconstruct or wait/sell. The developed model is illustrated and validated using data on damage and recovery of Staten Island, New York after the 2012 Hurricane Sandy. The results from simulation confirm that the combination of internal, interactive, and external drivers affect households’ decisions and shape the recovery. RecovUS is also applied to exemplify its potential application as a tool that could help practitioners and decisionmakers, who oversee allocation of mitigation and recovery resources and prioritization of restoration activities, predict impact of various factors, such as distribution of financial aids and recovery of infrastructure, homes and other community assets, on households’ recovery decisions. Like all studies, this research included limitations. A major part of the limitations was related to unavailability of micro-level data. Estimating individual-level damage and recovery of homes from tax assessment data, generating household-level attributes from census data, reimbursing financial resources for housing repair/reconstruction, and estimating damage and recovery of infrastructure and community assets from qualitative reports incorporated assumptions that although were based on literature, deserve to be future lines of study. Additionally, while RecovUS is neither underfit nor overfit the input data, verifying generalizability of its to different disasters, socioeconomic structures, distributions of financial resources, and restoration of infrastructure and community assets needs further research. Including more parameters for training if technical constraints permit, recollecting data to verify classification of households’ preferences for perceived neighborhood, exploring recovery of other occupancy types and physical characteristics such as rental, secondary, and multifamily properties, and simulating other aspects of recovery like business or infrastructure recovery are other important subjects that require more study. This research was supported by the National Science Foundation award #1454650: RecovUS - An Agent-Based Model of Collective Post Disaster Housing Recovery. Therefore, RecovUS was implemented based on procedures, assumptions, and data from the United States. However, the framework is general and could be easily modified to accommodate different procedures and conditions of other countries. RecovUS contributes to the domain of recovery modeling by proposing a model that simulates housing recovery decisions through mindset of insiders. In RecovUS, not only factors such as level of damage, financial resources, and affordability and availability of rental properties, but also interactions of households with their perceived neighborhood affect their decisions in favor of repairing/reconstructing their homes, waiting, or selling their properties. The modular structure of the model is another advantage that facilitates modification of the model without impacting its general integrity. Another feature is application of publicly available data to minimize the cost and time of data collection.

Description

Keywords

Disaster recovery, Recovery modeling, Agent-based modeling, Perceived community

Citation