Bias correction techniques for uncertainty reduction of long-term metocean data for ocean renewable energy systems

The design of effective and reliable solutions for Offshore Renewable Energy (ORE) technologies greatly relies on accurate metocean data. Uncertainties can significantly impact the design process. This paper presents a thorough evaluation of different bias-correction (BC) techniques applied to re-analysis datasets from diverse locations along the Spanish coast, presenting a multi-level evaluation procedure including novel, more appropriate statistical metrics beyond the traditional assessment techniques. First, the quality of raw ERA5 datasets is demonstrated to be rather poor, especially under extreme events, confirming the need for BC. Once the need is identified, results show that only the most sophisticated distribution-mapping BC techniques present the capacity to significantly improve the quality of the datasets, with the linearly-spaced Quantile-mapping (QM) showing the best overall performance for power production conditions (PP), reducing the bias by over an order of magnitude. In contrast, the Gumble Quantile-Mapping (GQM) outperforms the QM in survivability conditions (Surv). However, when computing overall performance of BC techniques, the predominance of PP hinders the relevance of Surv. Hence, adapted metrics with a more realistic balance between the PP and Surv regions are suggested, which show a better suitability of GQM providing more accurate estimations of the average power density and variability.