Marlo Armstrong | January 24th, 2025
With hurricanes hitting harder and more often, one question remains: why do some regions bounce back faster than others? Vanderbilt civil and environmental engineer Hiba Baroud collaborated with Sandia National Laboratories to provide new insights into what drives natural disaster-related power outages and their intensity. With the rise of tropical cyclones, this research plays a key role in figuring out the statistics behind which areas are more vulnerable to outages.
Dr. Baroud’s research published in March 2024 in Environmental Research Letters aimed to identify the primary factors influencing the duration of power outages and the number of people affected during tropical cyclones. Power outages are not only inconvenient but have serious implications for disaster response.
“Having electricity is extremely essential for disaster response and management,” Dr. Baroud said. “It’s critical for communicating warning and evacuation messages and functioning other essential services like transportation and water supply systems.”
The study focused on county-level data across the United States and used machine learning techniques to predict power outage risks. Dr. Baroud’s team analyzed four years of power outage data, working to link these outages with weather events using the National Oceanic and Atmospheric Administration (NOAA) database. Through this analysis, the research aimed to determine some of the underlying characteristics that may influence the length of county-wide power outages.
The study’s results show that weather variables, particularly conditions like wind speed and hurricane intensity, are the strongest predictors of power outages. This finding challenges previous assumptions that socioeconomic factors, such as income level or community resilience, were the key drivers of outage duration.
Vanderbilt risk and resilience analysis PhD student Celine Wehbe reflected on these findings.
“This study introduces a novel analysis by employing machine learning techniques to predict county-level power outage duration and the percentage of customers affected by tropical cyclones, incorporating nonlinear interactions among all variables,” Wehbe said.
The use of machine learning allowed the team to uncover complex, nonlinear relationships in the data that traditional models might have missed, providing a more accurate picture of how various factors interact during extreme weather events. However, it is important to note ways in which this study could be improved for the future.
“One limitation of this study is its focus on county-level data, which may overlook significant socioeconomic disparities present at more granular scales, such as neighborhood levels,” Wehbe said.
Dr. Baroud emphasized this limitation.
“At a higher resolution, like neighborhood-level data, socioeconomic variables may actually become more important,” Dr. Baroud said.
Wehbe says that this study advances the field by demonstrating that socioeconomic indicators may not significantly predict outage duration after controlling for weather conditions, challenging existing assumptions and encourages researchers to reevaluate the role of these variables. Additionally, there may be other variables that were not included in this analysis like infrastructure quality or local emergency response systems.
Looking forward, Dr. Baroud’s team hopes to improve the accuracy of their models by including more localized data and examining how climate change could affect future power outages. They are focusing on how different weather variables might change 50 to 100 years from now and supporting long-term planning as climate patterns evolve.
As the frequency and intensity of natural disasters continue to rise, these insights offer a valuable framework for future disaster preparation. While weather remains a dominant factor in predicting power outages during tropical cyclones, this study emphasizes that more research is needed at the local level to better understand how socioeconomic factors can inform more tailored disaster responses.
References
Johnson, P., Jackson, N., Baroud, H., and Staid, A. (2024) Can socio-economic indicators of vulnerability help predict spatial variations in the duration and severity of power outages due to tropical cyclones? Environmental Research Letters, https://iopscience.iop.org/article/10.1088/1748-9326/ad3568