QUANTIFICATION OF CHRONIC COASTAL FLOODING USING MACHINE LEARNING

Coastal communities are experiencing more frequent flooding outside of extreme events due to sea-level rise (SLR) (Hague et al., 2022; Sweet et al., 2021; Taherkhani et al., 2020). By 2100, 360 million people globally will occupy land at risk of flooding due to SLR (Kulp and Strauss, 2019). Although typically shallow in depth, these floods generate a chronic, cumulative impact on coastal communities by corroding infrastructure with saltwater (e.g., stormwater networks, roadways), reducing business revenues, and disrupting daily activities (Hino et al., 2019). These floods are difficult to measure with in situ sensor networks (e.g., pressure sensors and ultrasonic distance meters) because they are hyper-local and often short-lived: floods vary from small puddles to entire roadway widths, and last for minutes to several hours. In contrast to in situ sensors, subaerial cameras can capture the extent of flooding over larger areas (multiple residential blocks) at relatively low cost and therefore may provide a new mechanism for characterizing flood risk in areas with limited in situ flood measurements.