Predicting real-time roadway pluvial flood risk: A hybrid machine learning approach coupling a graph-based flood spreading model, historical vulnerabilities, and Waze data

Safaei-Moghadam, Arefeh; Hosseinzadeh, Azadeh; Minsker, Barbara

doi:10.1016/j.jhydrol.2024.131406

Urban flooding is a growing threat due to land use and climate change. Vulnerable populations tend to have greater exposure to flooding as a result of historical societal and institutional processes. Most flood vulnerability studies focus on a single large flood, neglecting the impact of small, frequent floods. Therefore, there is a need to investigate inequitable flood exposure across a range of event magnitudes and frequencies. To explore this question, we develop a novel score of inequitable flood risk by defining risk as a function of frequency, exposure, and vulnerability. This analysis combines high-resolution, parcel-scale compounded fluvial and pluvial flood data with census data at the census block group scale. We focus on six census tracts within Athens-Clarke County, Georgia that are highly developed with diverse populations. We define vulnerable populations as non-Hispanic Black, Hispanic, and households under the poverty level and use dasymetric mapping techniques to calculate the over-representation of these populations in flood zones. Inequitable risks at each census tract (approximately neighborhood scale) were estimated for multiple (e.g., 5-, 10-, 20-, 50-, and 100-year) flood return periods. Results show that the relatively greatest flood risk inequities occur for the 10-year flood and not at the largest event. We also found that the size of inequity is dynamic, depending on the flood magnitude. Therefore, addressing a range of events including smaller, more frequent floods can increase equity and reveal opportunities that may be missed if only one event is considered.

More Like this