Machine Learning Approaches for Meteotsunami Forecasting on the Coast of Portugal

We explored the capability of forecasting meteotsunamis using machine learning (ML) approaches. We selected meteotsunami events along the coast of Portugal where the atmospheric pressure jumps propagate from the south and southwest. Since this type of meteotsunamis is usually observed along the entire coast of Portugal (Kim & Omira, 2021; Kim et al., 2022), the southern tide gauges can act as a meteotsunami precursor for forecasting the northern coastal areas. For training and testing sets of ML, we started with the atmospheric pressure records (18 cases) which induced meteotsunamis, and then performed 1296 numerical simulation by varying the pressure inputs with different strength (jump magnitude), speed and direction. Then, the tidal gauge data from numerical simulations were used to apply neural networks (variational autoencoders and ARIMA) and to demonstrate the capability of meteotsunamis forecast based on one or more tide gauge observations. We observed that the ML models are capable of providing good predictions from short duration observations from the southern tide gauges. This work is supported by the project FAST—Development of new forecast skills for meteotsunamis on the Iberian shelf—ref. PTDC/CTAMET/32004/2017-funded by the Fundação para a Ciência e Tecnologia (FCT), Portugal.

 

References

Kim J, Omira R (2021) The 6–7 July 2010 meteotsunami along the coast of Portugal: insights from data analysis and numerical modelling. Nat Hazards 106:1397–1419. https://doi.org/10.1007/s11069-020-04335-8

Kim J, Omira R, Dutsch C (2022) Meteotsunamis along the Portugal coast from 2010 to 2019. 2nd World Conference of Meteotsunamis

Liu CM, Rim D, Baraldi R, LeVeque RJ (2021) Comparison of Machine Learning Approaches for Tsunami Forecasting from Sparse Observations. Pure Appl Geophys 178:5129–5153. https://doi.org/10.1007/s00024-021-02841-9

Omira R, Ramalho RS, Kim J, et al (2022) Global Tonga tsunami explained by a fast-moving atmospheric source. Nature 609:734–740. https://doi.org/10.1038/s41586-022-04926-4