ASSESSMENT AND MODELLING OF TSUNAMI MULTI- HAZARDS IN THE INDO-PACIFIC REGION

Abstract

The Indo-Pacific region is highly susceptible to tsunamis due to its unique geological characteristics. The area is characterized by major subduction zones where tectonic plate interactions can generate powerful earthquakes capable of triggering massive tsunami waves. The occurrence of frequent seismic activity along plate boundaries and fault lines further increases the risk of tsunami-generating events. This study focuses on advancing tsunami hazard assessment and forecasting in the eight identified tsunamigenic zones in the Indo-Pacific region, with particular emphasis on the Indian Ocean and its surrounding coastal zones. The research adopts an integrated approach combining stochastic techniques, artificial intelligence, and numerical modelling to improve the understanding of tsunami hazards, enhance forecasting accuracy, and evaluate their potential impacts on vulnerable coastal regions under worst-case scenarios. For this purpose, a comprehensive catalogue of historical and recent earthquake and tsunami data from 416 AD till 2022 for the Indo-Pacific region was compiled from various sources. This catalogue was then processed and filtered to ensure completeness in terms of time and tsunami intensity and then used for further analyses. To enhance the reliability of tsunami hazard assessment, this study develops regional magnitude correlation models for tsunamigenic zones across the Indo-Pacific region. These models focus on homogenizing magnitude scales by converting surface wave magnitudes (MS) to moment magnitudes (MW). Among the three regression approaches evaluated, Orthogonal Standard Regression (OSR) proved the most effective, providing more consistent and accurate estimates, especially in cases where independent and dependent variables exhibit considerable variability. The homogenisation relations developed in this study are tailored for the Indian Ocean region and its vicinity. They are highly recommended over existing global correlation equations for hazard estimation in regions lacking regional magnitude homogenization relations for tsunamigenic earthquakes. Further, for forecasting future tsunami hazards in the region, the research employs an integrated approach combining stochastic techniques and artificial neural networks (ANN). The stochastic approach involves fitting historically observed earthquake and tsunami recurrence times to several probability distribution models, including Gamma, Lognormal, Weibull, and Log-Logistic, and determining the most appropriate distribution model using the Bayesian Information Criterion (BIC). This analysis provides year-wise probabilities of occurrence for tsunami and tsunamigenic earthquake events in the Indo-Pacific region. To forecast the magnitude (Mw) of future tsunamigenic earthquakes, an ANN model was developed and trained with historical data on tsunamigenic activity, using 70% of the data for training. The performance of the ANN model was evaluated through mean square error (MSE) and the coefficient of determination (R²). The forecast results were validated using multiple approaches to ensure robustness and accuracy. The forecast results identify several high-risk zones for potential strong tsunamigenic events. Notably, Zones 5 (Tohoku Seismic Zone), 7 (Banda Arc), and 8 (Solomon Subduction Zone) are projected to experience strong events between 2030 and 2034, with magnitudes ranging from 7.0 to 8.2 Mw. The strongest tsunamigenic earthquake is forecasted for Zone 1 (Sumatra Subduction Zone), with an expected magnitude between 8.9 and 9.3 Mw, likely to occur between 2051 and 2064. Additionally, the study advocates that earthquakes with focal depths less than 80 km and MW ≥ 5.9 are likely to generate tsunamis in the Indo-Pacific region.