Parallel density scanned adaptive Kriging to improve local tsunami hazard assessment for coastal infrastructures

• Published in: Reliability engineering & system safety Vol. 222; p. 108441
• Main Authors: Di Maio, F., Belotti, M., Volpe, M., Selva, J., Zio, E.
• Format: Journal Article
• Language: English
• Published: Barking Elsevier Ltd 01.06.2022; Elsevier BV; Elsevier
• Subjects: Centroids; Clustering; Computer applications; Density; Earthquakes; Engineering Sciences; Hazard assessment; Hazard Curve; Oil & gas industry; Parallel density scanned Adaptive Kriging (P-ds AK); Refineries; Refinery; Reliability engineering; Seismic activity; Seismic probabilistic tsunami hazard assessment (SPTHA); Statistical analysis; Tsunamis; Wave height; Weather hazards; Seismic probabilistic tsunami hazard assessment (SPTHA); Parallel density scanned Adaptive Kriging (P-ds AK); Oil & gas industry; Hazard Curve; Refinery
• ISSN: 0951-8320; 1879-0836
• DOI: 10.1016/j.ress.2022.108441

•SPTHA must integrate the results of a large number of computationally demanding tsunami simulations.•Parallel density scanned adaptive kriging (P-ds AK) is used to address the computational challenge.•P-ds AK identifies the clusters centroids that most influence the hazard intensity (i.e., wave height).•The approach is applied in the area of the oil refinery located in Milazzo (Sicily, Italy). Seismic Probabilistic Tsunami Hazard Assessment (SPTHA) is a framework for calculating the probability that seismically induced tsunami waves exceed a specific threshold height, over a given time span and a specific region (i.e. regional SPTHA) or site (i.e. local SPTHA). To account for the uncertainty of the possible sources, SPTHA must integrate the results of a large number of computationally demanding tsunami simulations In this work, we innovatively use Parallel density scanned Adaptive Kriging (P-ds AK) to overcome the computational efficiency challenge of local SPTHA within a framework that consists in modeling/retrieving the full spectrum of possible earthquake triggering events at the regional level, filtering sources not relevant for the target, adopting a clustering procedure to select “representative scenarios” for inundation modeling, and, finally, adopt P-ds AK to identify the clusters centroids that most influence the hazard intensity (i.e., wave height) in the areas of interest. This approach is applied in the area of the oil refinery located in Milazzo (Italy). The application shows a consistent reduction of the number of high-resolution tsunami simulations required for the evaluation of the hazard curves over a set of inland Point of Interest (PoIs), either concentrated in one specific area or distributed along the coast.