HMN 2025: How Dataset reveals the factors affecting retail and charitable food supplies after Hurricane Harvey

Texas and Louisiana withstood the worst of Hurricane Harvey, which unleashed cataclysmic rain in August of 2017 that killed over 100 people from flooding. After the flood, grocery stores and pantries struggled not only to remain open but to keep fresh food on the shelves.

A new dataset that found key factors affecting post-Harvey retail and charitable food supplies has been awarded a 2025 NHERI DesignSafe Dataset Award, which recognizes the dataset’s diverse contributions to natural hazards research.

“This project explored the many dimensions of food access—supply chain, staffing, and infrastructure—focusing on Harris, Jefferson, and Orange counties in Southeast Texas, the three most heavily flooded by Hurricane Harvey,” said dataset principal investigator Nathanael Rosenheim, a research associate professor in the Department of Landscape Architecture and Urban Planning in the College of Architecture at Texas A&M University.

The award-winning dataset PRJ-2769 | Food Access Impact Survey for Harris County and Southeast Texas after Hurricane Harvey in 2017 was co-authored by Maria Watson of the University of Florida, and John Patrick Casellas Connors, Mastura Safayet, and Walter Peacock of Texas A&M University.

The dataset is publicly available on the NHERI DesignSafe cyberinfrastructure, which is managed at UT Austin and the Texas Advanced Computing Center (TACC).

Food retailers and food pantries experienced extensive building damage and electricity loss, which affected refrigeration and freezers, and caused staff shortages.

“We were interested in the link between food retail and food pantries, because after a disaster, food retail might see a decline because of disruptions, but food pantries will see an increase in demand,” Rosenheim said. “Disasters make those interdependencies more visible. We wanted to understand the weak points and help communities design more resilient food systems for both emergencies and during normal times.”

Data structure

The data curator for NHERI DesignSafe helped Rosenheim and colleagues develop the schema, or structure, for their award-winning dataset.

“It’s unique in the data archive world, because this project has interviews from 210 food retailers and 32 food aid agencies in terms of direct responses,” Rosenheim explained. “That information is geographically linked to the location of the food supplier. We also have additional data on other food retailers not interviewed across the region.”

A team of 15 graduate students rented cars to drive to Texas locations such as Orange, Beaumont, Port Arthur, and Houston for face-to-face interviews with as many stores as possible of the 500 in their sample of 3,000 retailers.

For each store in the tri-county area, the researchers conducted a nutrition environment survey, determining what products were available before and after Harvey.

“There’s a lot of variation—convenience stores don’t have as wide a selection of food as large grocery stores. All information is stored and archived in this project,” Rosenheim said.

Cyberinfrastructure aids food access modeling

The team used TACC’s Lonestar6 supercomputer to run models on the dataset with Python code, Jupyter notebooks, and the Stata statistics software.

“The tip of the data iceberg is the plain text CSV file. DesignSafe allowed us to store immense amounts of information,” Rosenheim said. “I use this in my teaching to demonstrate all the steps, all the assumptions, and all the decisions that were made over the several years since starting the project shortly after Hurricane Harvey in 2017.”

Outside of the classroom, Rosenheim sees the dataset as valuable to food systems planners, urban planners, emergency managers, and researchers.

“The results could help major metropolitan areas impacted by an event that affects a large part of that community,” Rosenheim said. “Emergency food planners could get a better idea of how long it takes for stores to recover from a hurricane. They would be able to be more strategic in how much, what type and where they send food resource assistance. The results will help model recovery timelines and identify vulnerable areas.”

Published results and bigger picture

Rosenheim and colleagues published findings from the research enabled by the dataset in the Journal of the American Planning Association in January 2024. They conclude that a focus only on store closures and property damage would underestimate the number of days residents had limited fresh food access by nearly 2 weeks.

The food access dataset is part of a larger project that developed a decision platform that integrates computational models of interdependent critical infrastructure systems such as transportation, energy, water, and food systems at different spatial and temporal scales.

DesignSafe continuity

“One of the great things about DesignSafe is that every person who contributed to this project is acknowledged. Having that metadata permanently stored in an archive is valuable, and many of the students associated with the project use it on their CV when they’re applying to graduate school or looking for jobs,” Rosenheim said.

Rosenheim stressed that post-disaster data is time-sensitive, and keeping track of the details of survey instruments and different types and versions of data can be challenging.

“DesignSafe has become a trusted place to keep track of what version of the data we used for a particular model. What’s more, the project has been active for years now. Having that kind of continuity of almost 10 years of archiving on DesignSafe has been amazing,” he added.

“Food access after disasters isn’t just about reopening stores, it’s about restoring true access, which depends on infrastructure, supply chains, and staff. That conclusion applies both to commercial retailers and to community-based food aid networks, like food pantries. The data helps identify where and why access breaks down, and it helps improve recovery planning.”

The content is provided for information purposes only.