RESUMO
Using volunteered geographic information (VGI) to supplement disaster risk management systems, including forecasting, risk assessment, and disaster recovery, is increasingly popular. This attention is driven by difficulties in detection and characterization of hazards, as well as the rise of VGI appropriate for characterizing specific forms of risk. Flash-flood historical records, especially those that are impact-based, are not comprehensive, leading to additional barriers for flash-flood research and applications. In this paper we develop a method for associating VGI flood reporting clusters against authoritative data. Using Hurricane Harvey as a case study, VGI reports are assimilated into a spatial analytic framework that derives spatial and temporal clustering parameters supported by associations between Waze's community-driven emergency operations center and authoritative reports. These parameters are then applied to find previously unreported likely flash flood-events. This study improves the understanding of the distribution of flash flooding during Hurricane Harvey and shows potential application to events in other areas where Waze data and reporting from official sources, such as the National Weather Service, are available.
Assuntos
Tempestades Ciclônicas , Desastres , Inundações , Medição de Risco , Tempo (Meteorologia)RESUMO
A real-time understanding of the distribution and duration of power outages after a major disaster is a precursor to minimizing their harmful consequences. Here, we develop an approach for using daily satellite nighttime lights data to create spatially disaggregated power outage estimates, tracking electricity restoration efforts after disasters strike. In contrast to existing utility data, these estimates are independent, open, and publicly-available, consistently measured across regions that may be serviced by several different power companies, and inclusive of distributed power supply (off-grid systems). We apply the methodology in Puerto Rico following Hurricane Maria, which caused the longest blackout in US history. Within all of the island's settlements, we track outages and recovery times, and link these measures to census-based demographic characteristics of residents. Our results show an 80% decrease in lights, in total, immediately after Hurricane Maria. During the recovery, a disproportionate share of long-duration power failures (> 120 days) occurred in rural municipalities (41% of rural municipalities vs. 29% of urban municipalities), and in the northern and eastern districts. Unexpectedly, we also identify large disparities in electricity recovery between neighborhoods within the same urban area, based primarily on the density of housing. For many urban areas, poor residents, the most vulnerable to increased mortality and morbidity risks from power losses, shouldered the longest outages because they lived in less dense, detached housing where electricity restoration lagged. The approach developed in this study demonstrates the potential of satellite-based estimates of power recovery to improve the real-time monitoring of disaster impacts, globally, at a spatial resolution that is actionable for the disaster response community.