Data-driven discovery of meteotsunami patterns from sparse observations

Abstract Meteotsunamis—tsunami-like sea level oscillations generated by atmospheric disturbances—pose underestimated risks to coastal regions worldwide. Despite growing evidence of their frequency and impact, limited offshore observations and forecasting capabilities hinder effective monitoring and early detection. Here, we present a data-driven framework for identifying and characterizing meteotsunami dynamics using sparse observational data. Leveraging dynamic mode decomposition and clustering techniques, we extract dominant spatiotemporal patterns and optimize the placement of offshore monitoring stations. We demonstrate the effectiveness of this approach using high-resolution simulations of the 2022 Ireland meteotsunami, a well-documented event exhibiting clear atmospheric forcing and sea-level response. Our results show that a minimal network of five strategically positioned sensors can accurately capture the essential dynamics of the event. This framework establishes a scalable methodology for designing cost-effective monitoring systems, enhancing our ability to detect and understand meteotsunamis under data-scarce conditions.