GNGTS 2024 - Atti del 42° Convegno Nazionale

Session 2.1 GNGTS 2024 Since the available inundation simulations were obtained only for the NEAMTHM18 (Basili et al., 2021) discretization, to not produce further simulation, we skipped the source refinement. The available simulations were sufficient to produce the target point refinement. A set of 8 target points were defined along the 10 m bathymetric line, covering the entire target area (see Fig. 2,a). The results of simulations were retrieved at these new points. The convergence of hazard curves with the N IS was checked also on these points (see Fig. 2,d-g). Focusing on the results of simulations on these new points, we conducted the cluster analysis for different numbers of clusters N CL (50, 100, 250): Fig. 2,b shows the WCSS applied to the N IS ensemble. For each of these cluster sets we have reproduced the hazard curves (N CL = 250 in Fig. 2,d-g) on the flow depth data on some points of the area of ​ interest and the inundation maps (N CL = 250 in Fig. 3), in order to compare the results obtained with those conducted starting from all the scenarios thanks to the HPC resources. As in Gibbons et al. (2020), no uncertainty is modelled in simulation results and a Heaviside step function is used that is 1 if the intensity computed by NLSW (Non-Linear Shallow Water) simulations for the scenario is greater than the reference MIH (Maximum Inundation Height) value, and 0 otherwise. Fig. 2: a) Setup for the cluster analyses in the Catania site. Subplot (a) shows a zoom of the finest grid for the Catania harbour from: in the wider map (b) it is highlighted the location of the target site on the Sicilian Island, with respect to which the disaggregation was conducted. (c) subplot displays the WCSS (Within-Cluster Sum of Squares) behaviour for an increasing number of clusters. (d), (e), (f), (g) subplots are the annual rates evaluated for the total (white curve), importance sampling (blue curve) and cluster (red curve) rates respectively: the plots refer to the same letter locations highlighted in (a) subplot. Hazard maps (for two different annual return periods) are displayed in Fig. 3: the differences between our results and the HPC ones are very small and nearly negligible for ARP 2500 yrs.