Byungho Kang, Rusty Feagin, Thomas Huff, Orencio Duran Vinent 

Low-intensity but high-frequency coastal flooding, also known as nuisance flooding, can negatively affect low-lying coastal communities with potentially large socioeconomic effects. Partially driven by wave runup, this type of flooding is difficult to predict due to the complexity of the processes involved. Here, we present the results of a probabilistic analysis of flooding events measured on an eroded beach at the Texas coast. A high-resolution time series of the flooded area was obtained from pictures using convolutional neural network (CNN)-based semantic segmentation methods, as described in the first part of this contribution. After defining flooding events using a peak-over-threshold method, we found that their size follows an exponential distribution. Furthermore, consecutive flooding events were uncorrelated at daily timescales but correlated at hourly timescales, as expected from tidal and day–night cycles. Our measurements confirm the broader findings of a recent multi-site investigation of the probabilistic structure of high-water events that used a semi-empirical formulation for wave runup. Indeed, we found a relatively good statistical agreement between our CNN-based empirical flooding data and predictions using total-water-level estimations. As a consequence, our work supports the validity of a relatively simple probabilistic model of high-frequency coastal flooding driven by wave runup that can be used in coastal risk management and landscape evolution models.