Flooding Beyond Limits: Capturing Unprecedented Surges and Flood Patterns in Large-Scale Coastal Flood Risk Assessments

Tropical and extratropical cyclones are among the most devastating natural hazards worldwide. By driving extreme storm tides, composed of storm surges and tides, they can cause severe flooding in low-lying coastal communities, resulting in loss of life and major economic damages. Model based coastal flood risk assessments are critical to reducing these impacts. At large scales, from continental to global levels, they can support organizations like the UN and EU in shaping disaster risk reduction policies, guiding investment decisions for international financing institutions, informing portfolios for (re-)insurance companies and helping multinationals descale risks. Historically, however, large-scale assessments have relied on many simplifications to remain computationally feasible. The coarse resolution of global coastal water level models and use of static flood modelling have led to inaccurate flood hazard estimates. Furthermore, extrapolating risk from short reanalysis datasets and assuming spatially homogeneous return periods along large coastlines often produced a misrepresentation of risk. The overall aim of this thesis is, therefore, to improve large-scale coastal flood risk modelling by better capturing realistic flood patterns and extremes beyond historical records. To advance coastal flood risk assessments, it is crucial to improve our understanding of global coastal flood hazards. In this thesis, this is achieved through the MOSAIC framework, which nests locally refined within coarser global models to generate high-resolution coastal water levels and simulate flooding hydrodynamically. MOSAIC was used in this thesis to assess the influence of model and output resolution, as well as bathymetry. To further improve coastal flood risk simulations, it is necessary to account for the spatial heterogeneity of risk. Building on MOSAIC, I developed a modelling framework to simulate stochastic coastal flood risk at large scales, reproducing realistic flooding events. A cascade of hydrodynamic models was applied to simulate 10,000 years of storm tides and flooding in Eastern Africa, followed by damage estimation and empirical derivation of country-level loss-exceedance curves. This approach provides information not available in traditional coastal flood risk assessments, such as event frequencies and magnitudes, the event contribution to annual damages and the ability to trace the coastal water levels, flooding and damages of each TC. In extratropical cyclone regions, longer timeseries than those available in current reanalysis-based datasets are necessary to reduce the uncertainties associated with modelling low-probability events. To address this , I created an extended storm tide dataset by pooling ensembles from SEAS5, coupled with GTSM, resulting in 525 years of storm tides. This dataset substantially reduces uncertainty in return period estimates compared to ~40-year reanalysis driven storm tide datasets, and provides more extreme events, thereby enhancing the reliability of extreme value distributions. Obtaining robust estimates of return periods is essential for many coastal flood risk decisions. However, understanding plausible yet unprecedented extremes is also required for certain management strategies. Using the 525-year storm tide dataset, I identified potential unprecedented storm surges in Europe and the Mediterranean coastlines beyond those recorded in the past 40 years, in terms of magnitude, spatial extent and temporal occurrence. Understanding unprecedented events across different dimensions provides crucial insights for coastal flood risk management. The research in this thesis contributes to ongoing efforts in the risk community to better understand coastal flood hazards and risks on a global scale. The MOSAIC framework enables high-resolution dynamic simulations of coastal water levels and flooding anywhere in the world, while the stochastic coastal flood risk modelling framework allows for more accurate large-scale risk estimation by incorporating the spatial dependencies of flooding events. Finally, the extended synthetic storm tide dataset provides more robust storm surge return periods and improves our understanding of unprecedented surges in Europe.