On the temporal sensitivity of the reference data in bias correction techniques for improving metocean datasets

Preprint submitted to RENEW 2024 conference. 

The paper presents a preliminary study limited to the use of wave height for the data corresponding to the Gulf of Biscay, a location with a very dominant North-West wave rose. It benchmarks different bias correction (BC) techniques and evaluates their performance, as well as the sensitivity of  the BC to the reference dataset employed for the identification of the BC parameters. More precisely, the amount of data (number of years) and the selected period (exact years) are analysed. Overall, the results demonstrate that the Gumbel-based BC techniques overperform the linearly-spaced BC techniques, the directional-adjusted Gumbel Quantile Mapping technique showing the lowest bias. With respect to the sensitivity of the reference data, BC seems to provide satisfactory results even when only 1 year of data are used. However, the dispersion among the different selected periods is large, resulting in large uncertainties. This dispersion reduces significantly once 3 or more years of data are used, independently of the selected period.