Analysis of tsunami impact scenarios at an oil refinery

This work analyses the potential consequences of two tsunami scenarios and their impacts on an oil refinery located in Sicily. Two credible tsunamis originating in the Tyrrhenian Sea were selected based on historical data. The potential for damage and hazardous materials releases resulting from the tsunami impacts to a refinery was assessed. The results obtained by the JRC tsunami propagation and inundation code HyFlux2 indicate that in both scenarios there would be eighteen storage tanks (of 43 located within 400 m from the shoreline) at the refinery subject to flooding. Water flow velocities were found to be generally low, <1 m/s, except for a central section of the refinery near the shoreline where the water flow velocities reach 3?C4 m/s. These results indicate that any damage would most likely occur due to buoyancy loads particularly in the western part of the facility where inundation levels are higher and storage tanks are less protected. Potential damage caused by impact of floating debris may be a problem in the central area of the refinery near the shoreline due to high flow velocities (3?C4 m/s) in both tsunami scenarios. Small hazardous materials releases could occur due to breakage of connected pipes and flanges caused by floating off of almost empty storage tanks or other equipment. Salt water intrusion could affect electrical equipment, such as control panels, pumps, and motors that are not raised above the inundation level. We conclude that in the two tsunami scenarios analysed, the risk to nearby residents and neighbouring facilities from potential hazardous materials releases, fires or explosions triggered by the tsunamis is likely to be small. Nonetheless, recommendations are made on prevention measures to reduce the risk of tsunami-triggered accidents and to mitigate their consequences if they do occur. The results of this study are limited by the uncertainty in the input data and most importantly by the accuracy of the elevation data and the model resolution.