Assessing future flood hazards with high-resolution climate model projections: a case study in northeastern Sicily, Italy

Flooding is one of the most frequent natural disasters with extensive impacts on society and the environment. Under projected climate conditions in the future, the occurrence of severe floods may increase. Therefore, assessing and increasing awareness of flood hazards is crucial for disaster risk reduction and adaptation planning. However, evaluating flood hazards under past and future extreme rainfall involves considerable uncertainties, as climate data at the required temporal and spatial resolution is most often missing. In this study, we use daily and sub-daily in situ observations (2001–2022) and projected hourly rainfall from an ensemble of high-resolution projections, exemplified for a coastal region in Sicily, Italy. We estimate the projected rainfall changes by using an ensemble of REMO2015 regional climate projections at a spatial resolution of 12.5 km and an hourly time-step, driven by 8 different CMIP5 general circulation models. We considered the historical climate (1981–2010) and the RCP8.5 scenario (2031–2060) to establish historical and near-future rainfall depth-duration-frequency (DDF) curves. With these data, we drive the hydrological model (HEC-HMS) for short-duration design rainfall events of 6 and 12 h in small urban coastal catchments in northeastern Sicily (Italy). The resulting flood depths, inundation area, and flow velocities were obtained from a 1D/2D hydrodynamic model (HEC-RAS) for 100-year design rainfalls. Our results show that future rainfall extremes are projected to be more frequent and severe in the study area, with the 100-year rainfall event increasing by 20% on average. This leads to increasing flood hazards compared to the baseline period, with inundation depth and velocity increasing by 9 and 12% on average in the study region. Our findings reveal that the regional climate model sub-daily rainfall data can reproduce extreme rainfall events obtained from observations despite some limitations, increasing confidence in future projections at this resolution.