Hindcast of Significant Wave Heights in Sheltered Basins Using Machine Learning and the Copernicus Database

Q: Why is long-term wave climate data important for maritime engineering?

Long-term wave parameter series underpin planning, operation, maintenance, coastal vulnerability assessment, and long-term coastal structure design using return levels and boundary conditions for models.

Q: What is the paper’s main goal for wave hindcasting?

To hindcast wave parameters using machine learning while relying only on publicly available Copernicus data, without requiring extended buoy measurement records.

Q: Which datasets and models are used, and how is the bias handled?

MEDSEA reanalysis wave parameters and ERA5 reanalysis wind parameters serve as inputs to ensemble regression and artificial neural networks; Bayesian optimization tunes hyperparameters, and the method substantially reduces bias in the original MEDSEA hindcasts.

Aug 3, 2026

Upload by Lucas Martin

Long-term wave-height time series are crucial for designing coastal structures and supporting maritime operations, yet buoy observations at many locations remain sparse. This study develops a machine-learning approach to hindcast significant wave heights and related parameters using only publicly available Copernicus products. Ensemble regression and artificial neural networks are trained with MEDSEA reanalysis wave fields and ERA5 atmospheric reanalysis wind inputs, with hyperparameters selected via Bayesian optimization. Results show up to 29% RMSE improvement for Rijeka and 12% for Split, alongside strong bias reduction to near-zero values.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Abstract

1. INTRODUCTION

Importance of long-term wave climate data

Limits of buoy measurements

Need for time-series extension

Hindcast of Significant Wave Heights in Sheltered Basins Using Machine Learning and the Copernicus Database Kratkoročna prognoza značajnih valnih visina u zaštićenim akvatorijima koristeći se strojnim učenjem i Copernicus bazom podataka Damjan Bujak* University of Zagreb Faculty of Civil Engineering [E-mail: damjan.bujak@grad.unizg.hr](E-mail: damjan.bujak@grad.unizg.hr) Tonko Bogovac University of Zagreb Faculty of Civil Engineering E-mail: [tonko.bogovac@grad.unizg.hr](tonko.bogovac@grad.unizg.hr) Dalibor Carević University of Zagreb Faculty of Civil Engineering [E-mail: dalibor.carevic@grad.unizg.hr](E-mail: dalibor.carevic@grad.unizg.hr) Tin Kulić University of Zagreb Faculty of Civil Engineering E-mail: [tin.kulic@grad.unizg.hr](tin.kulic@grad.unizg.hr) DOI 10. 17818/NM/2023/2 .5 UDK 551.466:004.85 004.032.2 Original scientific paper / Izvorniznanstveni rad Paper received / Rukopis primljen: 12. 7. 2022. Paper accepted / Rukopis prihvaćen: 20. 4. 2023. Abstract Long-term time series of wave parameters play a critical role in coastal structure design and maritime activities. At sites with limited buoy measurements, methods are used to extend the available time series data. To date, wave hindcasting research using machine learning methods has mainly focused on filling in missing buoy measurements or finding a mapping function between two nearshore buoy locations. This work aims to implement machine learning methods for hindcasting wave parameters using only publicly available Copernicus data. Ensemble regression and artificial neural networks were used as machine learning methods and the optimal hyperparameters were determined by the Bayesian optimization algorithm. As inputs, data from the MEDSEA reanalysis wave model were used for the wave parameters and data from the ERA5 atmospheric reanalysis model were used for the wind parameters. The results of this study show that the normalized RMSE of the test data improved by 29% for Rijeka and 12% for Split compared to the original MEDSEA wave hindcast at buoy locations.The proposed method was extremely efficient in removing bias in the original MEDSEA hindcasts (e.g., NBIAS = -0.35 for Rijeka) to negligible values for both Split and Rijeka (NBIAS < 0.03). Sažetak Dugoročne vremenske serije parametara vala igraju značajnu ulogu u projektiranju pomorskih građevina ipomorskim aktivnostima. Na mjestima s ograničenim direktnimmjerenjima plutača metode se koriste kako bi se proširio vremenski niz raspoloživihpodataka. Do danas, istraživanja o uspostavi kratkoročnih prognoza korištenjem metodama strojnog učenja uglavnom se usredotočilo na nadomještanje nedostajućihmjerenih podataka plutače ili pronalazak funkcijske veze između dvaju mjesta plutačau blizini obale. Ovaj rad ima za cilj koristiti metode strojnoga učenja radi provedbekratkoročnih prognoza koristeći se samo javnodostupnim podacima Copernicus. Ansambl regresija i umjetne neuralne mreže koriste se kao metode strojnoga učenja, aoptimalni hiperparametri određeni su Bayesianovim algoritmom optimizacije. Podaci valova iz MEDSEA modela i podaci vjetra iz ERA5 modela atmosfere su korišteni kaoulaznipodaci. Rezultati ove studije pokazuju da je za testni set RMSE se smanjio za 29% zaRijekui 12% za Split uspoređujućis izvornom MESEA kratkoročnom prognozom valovana lokacijama plutača. Predložena metoda bilaje izuzetno djelotvorna pri uklanjanju pristranosti u izvornoj MEDSEA kratkoročnoj prognozi (npr. = -0,35 za Rijeku) do zanemarivih vrijednostii za Split iza Rijeku (NBIAS < 0.03). KEYWORDS machine learning significant wave height ANN ensemble regression CMEMS KLJUČNERIJEČIstrojno učenjeznačajna visina vala ANN regresijasklopa CMMES 1. INTRODUCTION / Uvod Knowledge of a long-term time series of wave climate (e.g. significant wave height, peak wave period, etc.) at a location is essential for planning, operation, and maintenance of maritime activities [1], flood protection engineering desi

FAQs

Why is long-term wave climate data important for maritime engineering?

Long-term wave parameter series underpin planning, operation, maintenance, coastal vulnerability assessment, and long-term coastal structure design using return levels and boundary conditions for models.

What is the paper’s main goal for wave hindcasting?

To hindcast wave parameters using machine learning while relying only on publicly available Copernicus data, without requiring extended buoy measurement records.

Which datasets and models are used, and how is the bias handled?

MEDSEA reanalysis wave parameters and ERA5 reanalysis wind parameters serve as inputs to ensemble regression and artificial neural networks; Bayesian optimization tunes hyperparameters, and the method substantially reduces bias in the original MEDSEA hindcasts.

Aug 3, 2026

Upload by Lucas Martin