Storm surge hazard in the Mediterranean: a consistent hindcast-projections framework

Coastal regions worldwide are increasingly exposed to natural hazards, several of which are projected to intensify under global warming, amplifying risks to coastal communities, ecosystems, and critical infrastructure. Storm surges are a key driver of extreme sea levels and, when combined with other contributions to total water level, can trigger damaging coastal flooding. This PhD research develops a basinwide, high-resolution storm surge modeling framework for the Mediterranean Sea, with the dual aim of producing a long, homogeneous hindcast suitable for climatescale and extreme-value analyses, and generating an internally consistent ensemble of future projections to investigate potential climate-driven changes in surge hazard. First, a high spatial and temporal resolution storm surge model is implemented using Delft3D-FLOW over the entire Mediterranean domain. The model produces a hindcast spanning 1979–2023 that is forced by hourly high-resolution atmospheric fields, including 10-m wind and mean sea level pressure, enabling the representation of both synoptic and mesoscale meteorological features controlling surge generation in a semi-enclosed basin. Model performance is evaluated against surge residuals derived from tide gauge observations using standard statistical indicators and seasonally resolved diagnostics. Over the full hindcast period, the model reproduces observed variability with a mean Pearson correlation of approximately 0.6, with systematically higher skill during the storm-active autumn-winter months and reduced skill in summer, when surge magnitudes are small and local-scale processes become comparatively more influential. Model results are benchmarked against existing Mediterranean storm surge datasets from previous studies, showing broadlyn comparable skill. The representation of peak events relevant for coastal hazard is further examined. This is assessed through biases in extreme values between model results and observations, together with timing offsets and specific event-detection metrics, which demonstrate a satisfactory ability to reproduce the occurrence and key characteristics of surge extremes. Next, the influence of Atlantic open-boundary conditions and associated barometric forcing is also examined. The analysis demonstrates the crucial role of accounting for the Inverse Barometer Effect in improving the representation of Mediterranean sea level variability, with clear gains relative to configurations neglecting this contribution. The hindcast is then used to characterize basin-scale extremes through spatial diagnostics of maxima and return levels, and a focused case study of the October 2018 Ligurian storm illustrates the model’s capability to reproduce the timing, spatial footprint, and intensity of this high-impact surge event. Then, building on the validated historical reconstruction, the same modeling framework is applied to generate storm surge projections up to the end of the XXI century under a selected climate change scenario. A 17-member ensemble is produced using distinct EURO-CORDEX forcing combinations derived from Global and Regional Climate Models, covering a historical reference period (1970–2005) and a future period (2030–2100). Finally, to enable coherent comparisons with the hindcast and across the ensemble, projections are bias-adjusted using a dedicated methodology and analyzed in terms of changes in surge statistics and extremes, including return levels associated with frequent and rare events. Ensemble results indicate spatially heterogeneous and generally modest future changes in Mediterranean storm surge extremes, with no basin-wide uniform trend emerging across models. At the basin scale, ensemble-aggregated return level changes typically remain within ±5 cm for the selected return periods, whereas individual model combinations can yield locally larger increases or decreases, highlighting substantial inter-model spread. Inter-comparison further suggests that the projected spread is more strongly driven by the choice of the driving Global Climate Model than by the Regional Model, with simulations sharing the same global model often exhibiting more similar spatial patterns. Overall, the datasets and methods developed in this thesis provide a foundation for basin-scale assessments of present and future storm surge hazard in the Mediterranean, supporting both scientific investigations and coastal risk management applications.