- Title
- A data-driven approach to the fraction of broken waves
- Creator
- Stringari, Caio E.; Power, Hannah E.
- Relation
- Australasian Coasts & Ports 2019 Conference. Proceedings of Australasian Coasts & Ports 2019 Conference (Hobart, Tasmania 10-13 September, 2019) p. 361-366
- Relation
- https://www.proceedings.com/50955.html
- Publisher
- Engineers Australia
- Resource Type
- conference paper
- Date
- 2019
- Description
- The fraction of broken waves (Qb) is a key parameter for parametric surf zone models. It is via this variable that these models control the energy dissipation in the surf zone. Historically, Qb has been obtained using probability distribution functions (PDFs) of the wave height (p(H)). This paper describes an alternative, data-driven approach to obtaining the fraction of broken waves that is a significant improvement over the more traditional approaches. This new model is based on an ensemble of regression trees in which Qb is learnt directly from an extensive field dataset. The ensemble uses three input parameters that are often available to coastal engineers: offshore significant wave height (𝐻𝑚0∞), offshore peak wave period (𝑇𝑚01∞), and time-averaged relative water depths relative to the mean sea level (h/𝐻𝑚0∞), and predicts Qb at an averaged given relative water depth. The results indicate that the model can predict the depth-dependent variability of Qb with a high degree of accuracy (averaged r2 ≥ 0.95, averaged root mean square error ≤ 0.05, averaged mean absolute error ≤ 0.04) in virtually no computational time. When compared to three widely used Qb models that are derived from PDFs of the wave heights, the model developed here showed significant improvement with reductions in the errors (average error reduction of 25%) and significant improvement for r2-scores (average increase ≥ 30%). Although complex, the method developed here could be advantageous over the more traditional approach because of its high degree of precision and accuracy and because it does not depend on prior knowledge of p(H). In summary, the present model could be used as a replacement for the formulation of Qb in parametric wave models, which should result in better overall predictions, and thus, in better coastal management tools.
- Subject
- machine learning; decision-tree regression; surf zone; energy dissipation; broken waves; SDG 14; Sustainable Development Goals
- Identifier
- http://hdl.handle.net/1959.13/1435934
- Identifier
- uon:39866
- Identifier
- ISBN:9781925627237
- Language
- eng
- Reviewed
- Hits: 1396
- Visitors: 1383
- Downloads: 0
Collections
Goal 14: Life Below Water
| Thumbnail | File | Description | Size | Format |
|---|