Dataset for machine learning-based wave height error predictions around the Iberian Peninsula

This dataset was created under the frame of the Portuguese project DAMA - DAta assimilation and MAchine learning for improving wave forecast systems (https://doi.org/10.54499/2023.14738.PEX) that aims to investigate the suitability of machine learning techniques to predict the error of wave height forecast systems at specific locations. Afterwards, the forecast predictions can be adjusted to minimize discrepancies with observations. This dataset contains information from more than 30 stations along the Iberian Peninsula and has three main subcategories: wave parameter model predictions, wave and meteorological data observations and meteorological model predictions. Wave predictions were extracted from the AIB system of Puertos del Estado (open source). Here 13 variables are included. Wave and meteorological observations collected by buoys managed by Puertos del Estado e Instituto Hidrográfico. Data was extracted from the In Situ TAC portal of the Copernicus Marine Service. The number of avaialable variables varies based on the type of equipment and location. Meteorological predictions were produced by MeteoGaliza and they are freely avaialable. Six variables are included. In total, the dataset has up to 138 variables per station and and it covers from 2018-10-11 to 2025-03-31. The temporal frequency is not constant and for each day, data was collected at 00:00 (midnight), 06:00, and 12:00 (midday). Observations were assessed and outliers were removed.