Wave runup observations are important for coastal management providing data to validate predictive models of inundation frequencies and erosion rates, which are vital for assessing the vulnerability of coastal ecosystems and infrastructure. Automated algorithms to extract the instantaneous water line from video imagery struggle under dissipative conditions, where the presence of a seepage face and the lack of contrast between the sand and the swash impede proper extraction, requiring time-intensive data quality control or manual digitization. This study introduces two novel methods, based on color contrast (CC) and machine learning (ML). The CC method combines texture roughness — local entropy — with saturation. Images are first binarized using entropy values and then refined through noise reduction by binarization of the saturation channel. The ML method uses a convolutional neural network (CNN) informed by five channels: the grayscale intensity and its time gradient, the saturation channel, and the entropy and its time gradient. Both methods were validated against nine manually labeled, 80 min video time series. The CC method demonstrated strong agreement with manually digitized water lines (RMSE = 0.12 m, r=0.94 for the vertical runup time series; RMSE = 0.08 m, r=0.97 for the 2% runup exceedance (R_2%); and RMSE = 3.88 s, r=0.70 for the mean period (T_m−1,0)). The ML model compared well with the manually labeled time series (RMSE = 0.10 m, r=0.96 for the vertical runup time series; RMSE = 0.09 m, r=0.97 for R_2%; and RMSE = 3.51 s, r=0.79 for T_m−1,0). Furthermore, the computed R_2% values of both methods show a good agreement with the formula proposed by Stockdon et al. (2006) for extremely dissipative conditions, with RMSE-values lower than 0.13 m and correlations exceeding 0.70 for manual, CC, and ML estimates. While the CC method is deemed applicable for wave-by-wave analysis under similar dissipative conditions with a smooth seepage face and sufficient turbulent swash, the ML method still struggles with new, unseen data. However, it shows promise for a broader application and serves as a viable proof of concept. Together, these methods reduce the need for manual processing and enhance real-time coastal monitoring, contributing to more accurate predictive modeling of runup events and a better understanding of nearshore processes.

@en

Beach groundwater and nearshore hydrodynamic data were collected during a field experiment along two dissipative beach transects on Galveston Island, Texas, in the fall of 2023. The monitored beaches serve as nesting habitat for the critically endangered Kemp’s ridley sea turtle. Conditions ranged from calm to stormy, with two storms occurring during the experiment, inundating the entire beach up to the dune toe. Collected hydrodynamic data include readings from pressure loggers submerged in the foreshore and mounted in groundwater wells in the backshore, data from two wave buoys about 1.5 km offshore, and GoPro timestacks of the instantaneous waterline (wave runup). Other collected data include bathymetry and topography surveys, subsurface temperature and moisture content readings, and sediment characteristics. This comprehensive dataset can be used to (1) study relevant beach inundation and groundwater processes, including their effect on the local ecosystem (e.g., repeated flooding of sea turtle nests), (2) study the propagation of nearshore hydrodynamic processes into the beach matrix and groundwater table, and (3) validate existing beach groundwater models.@en