A Wave Data Assimilation System based on the Ensemble Kalman Filter (EnKF) is implemented for the North Sea showing improved performance and physical consistency. We first show the EnKF implementation and illustrate the wave data assimilation system using identical twin experiments to assimilate synthetic observations from buoys. A sensitivity analysis shows that the ensemble size, assimilation frequency and observation uncertainty are relatively important settings. Lastly, the potential for assimilating satellite measurements was assessed by assimilating synthetic altimeter measurements with real pass-over tracks. In these experiments, the state contains the full wave spectrum, unlike in most existing schemes. The results show that wave spectra and integral variables beyond significant wave height show physically consistent updates for the buoy and satellite experiments, by assimilating only significant wave height. This is a key advantage of this implementation compared to the more widely used implementations in wave data assimilation. Although the satellite experiment performs slightly worse than the buoy experiment due to decreased temporal availability of measurements, the results underline the potential for assimilation of satellite altimeter measurements. Such a system provides a promising framework for future observation impact study using satellite altimeter measurements.

A Bayesian network (BN) approach is used to model and predict shore-break-related injuries and rip-current drowning incidents based on detailed environmental conditions (wave, tide, weather, beach morphology) on the high-energy Gironde coast, southwest France. Six years (2011-2017) of boreal summer (15 June-15 September) surf zone injuries (SZIs) were analysed, comprising 442 (fatal and non-fatal) drownings caused by rip currents and 715 injuries caused by shore-break waves. Environmental conditions at the time of the SZIs were used to train two separate Bayesian networks (BNs), one for rip-current drownings and the other one for shore-break wave injuries. Each BN included two so-called "hidden"exposure and hazard variables, which are not observed yet interact with several of the observed (environmental) variables, which in turn limit the number of BN edges. Both BNs were tested for varying complexity using K-fold cross-validation based on multiple performance metrics. Results show a poor to fair predictive ability of the models according to the different metrics. Shore-break-related injuries appear more predictable than rip-current drowning incidents using the selected predictors within a BN, as the shore-break BN systematically performed better than the rip-current BN. Sensitivity and scenario analyses were performed to address the influence of environmental data variables and their interactions on exposure, hazard and resulting life risk. Most of our findings are in line with earlier SZI and physical hazard-based work; that is, more SZIs are observed for warm sunny days with light winds; long-period waves, with specifically more shore-break-related injuries at high tide and for steep beach profiles; and more rip-current drownings near low tide with near-shore-normal wave incidence and strongly alongshore non-uniform surf zone morphology. The BNs also provided fresh insight, showing that rip-current drowning risk is approximately equally distributed between exposure (variance reduction VrCombining double low line14.4ĝ€¯%) and hazard (VrCombining double low line17.4ĝ€¯%), while exposure of water user to shore-break waves is much more important (VrCombining double low line23.5ĝ€¯%) than the hazard (VrCombining double low line10.9ĝ€¯%). Large surf is found to decrease beachgoer exposure to shore-break hazard, while this is not observed for rip currents. Rapid change in tide elevation during days with large tidal range was also found to result in more drowning incidents. We advocate that such BNs, providing a better understanding of hazard, exposure and life risk, can be developed to improve public safety awareness campaigns, in parallel with the development of more skilful risk predictors to anticipate high-life-risk days.