Observations of beach change and runup, and the performance of empirical runup parameterizations during large storm events
Lauren Nicole Kim, Katherine Brodie, Nicholas Cohn, Sarah N. Giddings, M. A. Merrifield
Abstract
Timeseries observations of beach elevation change and wave runup from a tower-mounted stationary lidar are used to assess the skill (skill=1−NMSE, where NMSE is the normalized mean square error) of 2% runup exceedence (R2%) estimates (Stockdon et al., 2006) during four storm events at Duck, NC, USA. The runup parameterization requires specification of the foreshore beach slope, however beach slope is generally unknown during high energy events, and pre-storm estimates of beach slope are often used as a proxy. R2% hindcasts are computed using the observed time-varying beach slope and a static pre-storm beach profile. The time-varying beach slope yields R2% skill of 0.57 compared to the observed R2%. Similar skill scores are obtained using a tidally variable beach slope, derived from the static pre-storm beach profile, due in part to limited beach volume (−10.4 to 7.7 m3/m) and foreshore slope (0.04 – 0.09 m/m) changes during the four events. The skill drops to −1.0 using seasonal mean beach slopes. Even using the observed beach profiles, the skill reduces (−0.86) just after the peak of two of the storms due to the appearance of beach cusps (∼ 40 m wavelength) in the swash zone morphology. For one storm (a Nor’Easter), runup is underpredicted by up to 1.0 m at high tides following the storm peak when cusps are present, which coincides with high tides of near equal magnitude to those near the storm peak. Additional pre- and post-storm mobile lidar surveys for one of the storms confirm ubiquitous small-scale beach cusps along 8 km of the local shoreline. The results suggest that skillful runup estimates during storms are often attainable given the availability of beach information just before a storm. The parameterization errors increase when beach cusps develop, highlighting the need to extend standard one-dimensional runup parameterizations to account for two-dimensional effects.