Remote Sensing and Qualitative Understanding (RES-Q): A mixed-methods analysis of flooding impact on unsheltered populations

Abstract Submission: In 2023, approximately 653,100 people were unhoused on a given night, of which 40% were experiencing unsheltered homelessness. This group is particularly susceptible to the escalating frequency and severity of natural disasters like flooding – a trend worsened by climate change. Existing flood modeling techniques fail to capture the unique situational vulnerabilities of unhoused individuals, making it difficult for government agencies to provide timely communications and aid, leading to loss of property and life. Here, we utilize a novel mixed-methods approach to quantify the differential flood vulnerability of areas with high populations of unsheltered people, using Los Angeles County as an initial case study for our model. We utilize machine learning-based flood prediction using remote sensing data to identify key areas of urban flooding. By conducting semi-structured interviews with community organizers and non-government organizations currently working with unhoused individuals in Los Angeles County, we account for place-based knowledge of the most affected areas and existing barriers faced by unhoused individuals. We consider how sociodemographic factors such as age, race and disability lead to higher individual risk during a flood event. Combining insights from the qualitative interviews with ML model outputs, we use the Analytical Hierarchy Process to identify the most vulnerable encampments.

These insights will empower our community partners in Los Angeles to make data-driven decisions about aid allocation, communication, and shelter capacitys. By blending data-driven geospatial modeling with qualitative analysis, our approach fosters a more inclusive and nuanced understanding of disaster vulnerability through collaborative engagement with diverse stakeholders.

Learning Objectives/Expected Outcome (Optional) :