Climate change is increasing the frequency and intensity of extreme weather events, causing life-threatening situations and critical infrastructure damage. These events disproportionately affect vulnerable populations (low socioeconomic status, minorities, elderly). Understanding these disparate impacts is crucial for developing equitable mitigation and response strategies. Previous research has shown inequalities in disaster impacts based on race, class, and other factors, particularly following events like Hurricane Katrina and Hurricane Harvey. However, studies on extreme cold weather are scarce. The Texas Winter Storm Uri in 2021, causing widespread power outages and significant economic losses, presents a unique opportunity to examine these inequalities. Managed power outages are increasingly common during extreme weather, but their equitable implementation is understudied. This study addresses this gap by using digital trace data to assess power outage impacts and other community-scale data to assess impacts on vulnerable populations in Harris County, Texas.

Existing literature demonstrates the disproportionate impact of disasters on vulnerable populations. Studies of Hurricane Katrina and Hurricane Harvey highlighted disparities in recovery based on race and socioeconomic status. Previous research also showed that less advantaged households experienced decreased ability to withstand service disruptions. However, studies focusing on extreme cold weather events are limited. The 2014 winter storm in Georgia and South Carolina and the 2021 Winter Storm Uri in Texas caused significant damage and loss of life, providing opportunities to study these events, yet detailed data remains limited.

This study used community-scale big data from multiple sources in Harris County, Texas, to assess the impacts of Winter Storm Uri (February 13-17, 2021). Data included: 1. **Population Activity:** Mapbox data provided telemetry-based population activity, used as a proxy for power outages. Activity density (Da) was calculated at the census-tract level by aggregating activity indices from a 100m x 100m grid. Data from January 11-17, 2021, served as a baseline. 2. **Point-of-Interest (POI) Visit Data:** SafeGraph data on visits to grocery stores and restaurants provided insights into food accessibility. Data were aggregated to the census-tract level using a one-mile buffer. 3. **311 Service Helpline Data:** Houston's 311 system data were used to assess burst pipes. Water-related service requests were filtered and aggregated to the census-tract level. 4. **Demographic Data:** The 2019 American Community Survey 5-year estimates provided data on income, race, and ethnicity at the census-tract level. High/low income and minority/non-minority groups were identified based on the top and bottom 25% of each respective category. **Statistical Analysis:** Kruskal-Wallis tests were used to compare impacts across income, race, and ethnicity groups due to the non-normality of the data. Spatial autocorrelation (Global Moran's I) was used to analyze spatial dependency in the results. Agglomerative hierarchical clustering was used to classify POI visit trends and identify patterns of responses to the storm.

The study found significant disparities in the impact of Winter Storm Uri across different subpopulations: **Power Outages:** * Low-income populations experienced significantly greater impacts in terms of the extent of power outages (greatest negative change in activity density). * Ethnic minority populations (specifically Hispanic) experienced significantly longer recovery times from power outages. * Significantly impacted census tracts (those with near total loss of activity) had lower median household incomes and higher percentages of Black and Hispanic residents than the county average. **Burst Pipes:** * Low-income and racial minority groups had significantly higher normalized 311 calls for water-related issues (burst pipes) per person per area compared to high-income and racial nonminority groups. * Twice as many high-income and racial nonminority census tracts were unaffected by burst pipes than low-income and racial minority groups. **Food Inaccessibility:** * The most impacted census tracts (Class a) regarding restaurant and grocery store visits were in Houston's downtown, which experienced decreased activity, but the secondary impacted census tracts (class b) showed lower median incomes and higher percentages of Black and Hispanic populations indicating food access issues among these demographics. * Spatial autocorrelation analysis showed that the most impacted areas for both grocery stores and restaurants were clustered near Houston's downtown area, indicating spatial patterns in food inaccessibility.

The findings demonstrate significant inequalities in the experience of Winter Storm Uri across different socioeconomic and racial/ethnic groups in Harris County. The use of community-scale big data provided valuable proxy measures for impact assessment, particularly considering the unavailability of fine-grained power outage data. The study highlights the disproportionate impact on vulnerable populations due to power outages, burst pipes, and food inaccessibility, reinforcing the need to incorporate social vulnerability into disaster preparedness and response strategies. The results underscore the necessity of equitable resource allocation and mitigation planning in the face of extreme weather events.

This study highlights the disparate impacts of Winter Storm Uri on low-income and minority populations in Harris County, Texas, showcasing the value of community-scale big data for rapid impact assessment. The findings reveal inequalities in the experience of power outages, burst pipes, and food inaccessibility. This underscores the need for infrastructure operators and policymakers to consider social vulnerability in disaster preparedness and response planning. Future research could focus on examining the causal mechanisms behind these disparities and developing more effective and equitable mitigation and adaptation strategies.

The study relied on proxy measures for power outages due to the lack of granular power outage data. While population activity data provided valuable insights, it might not perfectly capture the full extent of power outage impacts. The aggregation of data to the census-tract level could also mask finer-grained variations within census tracts. The Modifiable Areal Unit Problem (MAUP) was not explicitly addressed and could be a focus for future research. Additionally, the study did not directly explore the causal mechanisms underlying the observed disparities.