PSI - Issue 78

Yazdan Almasi et al. / Procedia Structural Integrity 78 (2026) 433–440

438

the sequential occurrence of an earthquake followed by a tsunami presents a compound risk scenario known to cause catastrophic consequences, as evidenced by the 2011 Tōhoku Earthquake. One of the primary challenges in multi hazard Na-Tech risk assessment lies in characterizing the sequential nature of the earthquake-tsunami scenario. Earthquakes may inflict damage to structural components such as ASTs through buckling, sliding, and anchorage failure, weakening their capacity to withstand the hydrodynamic forces of a subsequent tsunami (Goda & De Risi, 2018; Vitale, Baltzopoulos, et al., 2024; Wu et al., 2024). This cascading mechanism is not purely additive—rather, it introduces path dependency in the system’s response, necessitating methods that account for inter-hazard correlations. Vitale, Baltzopoulos et al. (2024) and Wu et al. (2024) explicitly acknowledge the importance of considering residual seismic damage before tsunami impact. Their approach integrates separate fragility functions for ground shaking and tsunami inundation but within a unified probabilistic framework, where both hazards are treated on the same site using harmonized hazard curves derived from consistent seismic source models. However, such approaches typically assess risk independently and only compare failure rates ex post, lacking dynamic interaction modeling between hazards. More advanced methodologies attempt to simulate this coupling through sequential fragility analysis. For instance, Xu et al. (2021) proposed a sequential fragility analysis framework in which the residual structural capacity after earthquake loading serves as the initial condition for subsequent tsunami analysis. Although this approach increases computational demand, it constitutes a significant advancement toward more realistic modeling of cascading hazard effects. Similarly, empirical studies by Nishino and Imazu (2018) on post-tsunami oil fires in Kesennuma revealed that structural damage patterns were indeed consistent with a sequential mechanism where ground shaking primed failure paths for tsunami-induced ignition and release. Probabilistic Tsunami Hazard Analysis (PTHA) and Probabilistic Seismic Hazard Analysis (PSHA) are well-established individually, yet integrating them remains methodologically fragmented. Dual-hazard models such as the one proposed by Rahimi et al. (2025) for the Makran coast adopt a risk-targeted approach, aligning exceedance probabilities for both hazards using logic trees and joint hazard space. Others, such as Goda and De Risi (2018), have developed stochastic earthquake-tsunami scenarios based on shared rupture models, enabling the construction of joint hazard curves from co-dependent simulations. However, most current risk assessments still rely on parallel processing of hazard curves and then combine failure probabilities using simplifications such as weighted averaging, worst-case assumption, or conditional exceedance. These methods lack a mechanistic basis and may either overestimate or underestimate risk due to unmodeled dependencies (Choi et al., 2024; Goda & De Risi, 2018; Vitale, Baltzopoulos, et al., 2024). A notable step forward is the concept of hazard convolution, where tsunami loading is sampled conditionally based on the preceding earthquake’s rupture properties—an approach suggested by Goda et al. (2014) and furthered in Vitale, Baltzopoulos et al. (2024) by applying matched hazard scenarios. The empirical basis for earthquake-tsunami interaction is steadily improving. Post-disaster reconnaissance studies from the Tōhoku 2011 and Kesennuma (Nishino & Imazu, 2018) events reveal a clear link between pre-tsunami structural damage and subsequent Na-Tech failures, such as oil spill fires and flammable gas releases. Historical events such as the 1908 Messina earthquake and tsunami also provide qualitative evidence of cascading hazard phenomena, even if their physical modeling remains less robust. Nevertheless, most empirical data are limited to a few case studies, with scarce quantitative data on sequential tank failure. There is thus a pressing need to develop observational fragility databases to validate numerical or probabilistic models under compound loading. Despite these advancements, several critical gaps remain. First, the scarcity of empirical data on sequential failure, particularly for large industrial tanks, limits the calibration of coupled fragility functions (Vitale, Baltzopoulos, et al., 2024; Wu et al., 2024). Experimental efforts such as those by Wu et al. (2024) remain rare and often focus on idealized conditions. Secondly, modeling frameworks often ignore the time delay between events, which could be minutes to hours, influencing recovery, emergency response, and water ingress dynamics (Choi et al., 2024; Vitale, Baltzopoulos, et al., 2024; Wu et al., 2024). Moreover, many industrial risk assessments still treat earthquake and tsunami as statistically independent, despite their physical correlation. The COHRISK framework (Choi et al., 2024) partially addresses this issue by allowing sequential relationships within its logic model; however, its application is currently limited to nuclear plants. General-purpose tools like CAPRA, HAZUS, or NARSIS-MHE focus on multi-hazard occurrence but lack structural system modeling capabilities, making them less suitable for industrial risk evaluation (Choi et al., 2024). Another challenge lies in system-level resilience modeling. The Markov recovery framework proposed by Dhulipala et al. (2021) offers a promising pathway by using state-dependent transitions to forecast post-event functionality under single or multiple hazards. This is particularly relevant for interconnected systems such as