Simulating extreme storm floods in the German Bight – past variability & future changes

Extreme high sea levels (EHSL) caused by storm floods are among the major hazards for low-lying coastal environments such as the German Bight. For adequate flood protection measures both now and in the future, coastal planners usually require estimates of return periods and magnitudes. Such numbers are subject to considerable uncertainties though, as they are typically obtained from statistical methods based on the relatively limited observational record which often spans a couple of decades only. Likewise, their potential future change, e.g. until the end of the century, is often investigated with dynamical simulations comparing two 30-year periods at beginning and end of the simulation period. Yet, how robust are such estimates really, given the often unknown internal variability on multidecadal to centennial time scales?

Building on long-term climate simulations from a regionally coupled climate system model focusing on the North Sea, we here argue that the above methods cannot reflect the full internal EHSL variability as they do not account for their long-term variations. The simulation of extreme sea level variations during the past millennium (Lang & Mikolajewicz, 2019) has shown that while the statistics compare well with the observational record, EHSL and especially the ‘high-impact-low-probability’ events vary substantially on interannual to centennial timescales.

This high EHSL variability has implications for the assessment of future changes in EHSL statistics. Given their large variations, such changes cannot be inferred from single simulations or small ensembles. Existing estimates of future EHSL changes from small samples are thus likely to be dominated by internal variability rather than climate change signals. Thus, large ensemble simulations are required to assess future flood risks.

Here we use 32 members of the ‘1pctCO2’ simulations of the MPI Grand Ensemble (Maher et al., 2019) as global forcing for an ensemble downscaling with the regionally coupled climate system model. Such a large ensemble allows the detection of changes even in high return values with low statistical uncertainty and may thus help to better assess future changes in storm flood statistics.