A secular sea level hindcast (1900–2015) to investigate extreme surges variability and trends in the North Atlantic

Changes in extreme sea levels, combined with the growth of coastal population, are critical factors in evaluating the risks related to coastal flooding. Thus, studying the variability and trends of storm surges, a major contributor to extreme sea levels, becomes essential for coastal protection policies. We developed in the North Atlantic the first hourly surge hindcast covering the full 20th century (1900–2015) on a 0.1°grid, and called ClimEx hindcast. We validated the hindcast against 34 long-term tide gauges. The model shows overall very good performance for surges (Root Mean Square Error of 9.3 cm on average), and good performance for extreme surges, despite an overall underestimation. To investigate the variability and trends in storm surges, we performed a non-stationary extreme value analysis on modeled and observed storm surges. The seasonality of storm surges is highly dependent on the area. The seasonal amplitude varies from typically 10 cm, to more than 40 cm in the North Sea. The storm surge season occurs around December–January in the north of the domain (above 40°N), due to winter extra-tropical cyclones, and around September–October in the south-west, due to tropical cyclones. The dependence of storm surges with the North Atlantic Oscillation extends from the coasts to the deep ocean, and is positive above 50°N and negative below. Observed storm surges show mostly non significant or small trends ($<\pm$ 1 mm/yr), while the model displays positive trends almost everywhere, possibly due to inhomogeneities in the atmospheric forcing dataset prior to 1950.