This work assesses the performance of different spatial interpolation methods (inverse distance weighting/IDW family, nearest neighbour, linear interpolation, natural neighbour, and spherical kriging) for the preliminary wind energy assessment and the estimation of wind speed and direction from numerical models that constitute key factors in early-stage feasibility studies for offshore wind farm (OWF) development. Using the CERRA reanalysis dataset over the Mediterranean Sea, long-term measurements from 31 buoys have been used as ground truth data, and the methods’ performance was evaluated through multiple statistical metrics and a weighted aggregated performance metric (WAPM). To ensure statistically robust comparisons, the non-parametric Friedman and Nemenyi tests were applied, along with the Aligned Rank Transform ANOVA to examine interactions between performance and distance from shore. The numerical results suggest that for wind power density and energy production, inverse distance weighted regression (IDW-R) and natural neighbour perform better than the rest of the interpolation methods and should be considered for assessing wind energy characteristics of candidate areas for OWF development. The same methods perform best for wind speed interpolation, while IDW-R and IDW0 (mean of four) perform best for wind direction. One of the most important advantages of the IDW-R is that it reduces local bias and improves accuracy due to the embedded linear regression framework, while its interpolation quality is superior when the available data points are limited. Overall, the numerical results clearly suggest that the selection of an appropriate interpolation method can significantly reduce errors in the preliminary estimation of the available wind power and projected offshore energy production.
Soukissian et al. (Thu,) studied this question.