Advanced weather typing for downscaling of wave climate and storm surge at a UK nuclear power station

Abstract Evaluating risks from external hazards is crucial for the safety of nuclear power stations throughout their lifecycle. In coastal areas, a key threat arises from the risks of coastal flooding and erosion via a combination of simultaneous processes (e.g., tides, waves, and storm surges) acting on varying spatial and temporal scales. Therefore, an accurate characterisation of local sea state conditions is essential for risk assessment and mitigation. In this paper, we use a weather typing method to downscale local wave climate and storm surge conditions at the Hartlepool nuclear power station. Model validation suggests that the use of 36 weather types can effectively downscale multivariate wave variables (wave height, period, and direction) and storm surge with overall good performance, though the accuracy is limited for wave direction and extreme wave height. Comprehensive sensitivity tests are conducted to investigate key factors influencing the downscaling process, including predictor variable, spatial and temporal definitions, predictor resolution, the number of weather types, and the weighting parameter in semi-supervised classification. For example, we find that the model with sea level pressure and sea level pressure gradient as the predictor has better overall performance in downscaling multivariate predictands than the model using either one individually. These results can facilitate the development of weather typing models to enable efficient and reliable estimations of local predictands in wider applications. This approach links atmospheric conditions to potential coastal threats, which offers a valuable tool for proactive hazard preparedness and risk management in nuclear power and other critical infrastructure sectors.