Predicting Wave Height through Wave Spectra Partitions and Machine Learning

The practical and sustainable deployment of coastal and marine operations, including shipping, marine energy, fishing, offshore exploration, and coastal infrastructure development, relies heavily on accurate and reliable ocean forecasting. The current state-of-the-art in ocean forecasting lies in the third-generation numerical wave models, such as ECWAM [1] and WAVEWATCH III [2]. For all their strengths and accomplishments, the numerical wave models are not perfect. The numerical models output directional wave spectra, which is a function that takes wave propagation direction and wave frequency as inputs and returns mean surface elevation variance as an output. However, the wave spectra contain various independent wave systems. Treating a combination of independent wave systems as a single, homogeneous whole makes its dynamics appear more convoluted than they are. Since (Gerling, 1991) partitioning has been used to separate the multitude of wave systems present in each wave spectrum.

Based on the idea of wave spectra partitioning, we propose two novel machine learning models, each accounting for a different modality of data. The first model will make use of summarising statistics of the partitioned wave systems by learning their time evolutions, which we will use to predict future significant wave heights. The second model will instead use the directional wave spectra in their entirety, once again learning the dynamics of the system. Both of our machine learning models carry different advantages; while the first model is less computationally demanding, the second might be more performant. We analyse both models and compare them to determine their strengths and weaknesses.

For about the project contact Mr Merlijn Surtel, Research Fellow at the Department of Mathematics and Statistics at the University of Strathclyde.

For a list of the research areas in which ARCHIE-WeSt users are active please click here.

References

[1]    ECMWF, “IFS Documentation– Part VII: ECMWF Wave Model,” Nov. 2016. Accessed: Jul. 09, 2024. [Online]. Available: https://www.ecmwf.int/sites/default/files/elibrary/2016/17120-part-vii-ecmwf-wave-model.pdf

[2]    H. Tolman, “User manual and system documentation of WAVEWATCH III TM version 3.14,” NOAA, May 2009. Accessed: Jul. 09, 2024. [Online]. Available: https://polar.ncep.noaa.gov/mmab/papers/tn276/MMAB_276.pdf