Risk of biases in estimating storm surge extremes: a global study

Extreme sea levels (ESLs) threaten coastal zones, and global assessments suggest that future ESLs grow with rising mean sea level (MSL) as the main driver. Conventional practice assumes that present-day return levels are added to a given MSL trajectory into the future to obtain future return levels. Here, we test this additive assumption and show that it may under- or overestimate ESLs with consequences for coastal protection decisions. We fit stationary and non-stationary generalised extreme value (GEV) models to annual maxima of long hourly tide-gauge records. The non-stationary GEV location parameter may depend on MSL, time, or both. Bayesian model comparison and posterior coefficient support classify 200 of the 386 analysed series (52%) as non-stationary and 186 as stationary. At some stations, ESLs increase more than MSL; at others they increase less than MSL or even decrease. Post-fit grouping by seabed depth and slope also suggests stronger morphology-related differences in the fitted MSL coefficient among non-stationary series than among the stationary comparison series.