UC San Diego

The accuracy of flood forecasting models depends crucially on understanding wave runup. I use theory, insitu and remote observations, numerical modeling, computer vision, and deep learning to (1) investigate numerically the runup dependence on bathymetry and incident wave conditions, (2) improve video-based bathymetry estimates, and (3) characterize infragravity waves in 10m depth, for use in boundary conditions of runup models. Implementation into operational runup observing systems and models is ongoing.

A numerical modeling (SWASH) study used 138 hindcast historical storm waves, two offshore boundary conditions, and 24 representative eroded beach bathymetries from a Southern California beach. The runup 2% exceedance level varied by more than 30% in response to changes in bathymetry or infragravity wave boundary conditions. An empirical parameterization trained on this dataset includes both a foreshore beach slope beta_f and an effective mid-surfzone slope beta_eff (Chapter 2, Lange et al. 2022). Subaqueous bathymetry is usually unknown because of the large expense of insitu jetski surveys, but beta_f and beta_eff can be estimated approximately and cost-effectively from images.

The new 2-slope runup models have smaller errors than 1-slope models, but lack generality and have fundamentally limited accuracy. I show that useful bathymetry can be extracted from video collected during a single 17-minute quadcopter hover. The existing cBathy algorithm is extended with a crest-tracking algorithm that significantly reduces large cBathy errors near the breakpoint. The crest-tracking algorithm uses a deep-learning neural network to annotate timestacks for celerity estimates, and the depth inversion includes a nonlinear correction. This approach reduces RMSE surfzone depth errors to 0.17m, from ~0.81m with cBathy (Chapter 3, Lange et al. 2023, revision submitted to Coastal Engineering).

The infragravity offshore (~10m depth) boundary condition is another potential error source in the runup model estimates. Several years of observations show that free infragravity waves are often much larger (up to x10) than the bound waves often used as a boundary condition. A parameterization of the incident-free IG wave field is combined with the predicted boundwave energy into a sea surface elevation timeseries of the incident IG energy suitable for use in numerical models (Chapter 4).