Due to the increasing interest in the prospection of potential sites for the installation of offshore wind farms, it becomes important to extend the tests presented on Carvalho et al. (2014) to offshore areas. For that, the WRF model was used to conduct ocean surface wind simulations forced by different initial and boundary conditions (NCEP-R2, ERA-Interim, NCEP-CFSR, NASA-MERRA, NCEP-FNL and NCEP-GFS) aiming to assess which one of these datasets provides the most accurate ocean surface wind simulation and offshore wind energy estimates. Six near surface wind simulations were performed, each one of them forced by a different initial and boundary dataset. Results were evaluated using data collected at five buoys that measure the wind in the Iberian Peninsula region (Galician coast and Gulf of Cádiz).

The results show that the simulation driven with ERA-Interim reanalysis provided the lowest errors in terms of offshore wind temporal variability. NCEP-R2 driven simulation showed the lowest offshore wind speed bias, mean wind speed and offshore wind energy production estimates. However, it was the one with the highest errors related to the wind temporal variability. The simulations driven with the NCEP-FNL and NCEP-GFS analyses products also showed interesting results, better than the NCEP-CFSR and NASA-MERRA reanalyses.

Based on the results presented in this work and in Carvalho et al. (2014), ERA-Interim reanalysis likely provide the most accurate initial and boundary data to force near-surface wind simulations for the offshore and onshore areas. However, for offshore sites the NCEP-R2 reanalysis seem to provide the most accurate estimation of the potential wind energy production, fact that is of great importance for the wind energy industry. Furthermore, the NCEP-GFS and NCEP-FNL analyses can be considered as valid alternatives to ERA-Interim and NCEP-R2, in particular for cases where reliable forcing data is needed for real-time applications due to their fast availability.