Key points are not available for this paper at this time.
Coastal floods are one of the main natural hazards which often lead to disasters worldwide including in Sweden. In the Baltic Sea, extreme sea levels (ESL) are driven by different processes such as seiches, pre-conditioning of the Baltic Sea (filling) and storm surges. Predicting those along with accurate uncertainties range is essential for marine coastal management and planning to better mitigate potential impacts. It is well known that Climate Change is currently changing coastal conditions globally and around Sweden. In this context, this study aims to characterize the atmospheric drivers of ESL along the coastline of Sweden and predict return levels of extreme sea heights. Machine Learning tools like Random Forest are then used to predict storm tide time series, characterising extreme sea levels, based on atmospheric drivers such as wind and surface pressure. The model is firstly trained with ERA5 reanalysis and tide gauge observations. It is then used to predict storm tides from 1850 to 2100 using input variables from CMIP6 models as EC-Earth. Predictions are then compared with observation values in a statistical way over the co-occurring time period. Results show that storm tides are driven by different atmospheric characteristics (i.e. wind direction, wind speed and surface pressure) at each station but that stations located close to each other can be grouped accordingly. For instance, strong westerly winds greatly contribute to storm tides on the west coast of Sweden while southernly winds are driving storm tides on the northern Swedish Baltic coast. Our main results include investigating trends in extreme values of storm tides. Based on the EC-Earth historical and ssp245 climate model, a decrease in storm tide of 1 to 4 mm/decade is to be expected along the east coast of Sweden while no trend seems to be noticeable on the west coast of Sweden.
Dubois et al. (Fri,) studied this question.