The ARPEGE and MASS models predicted higher 10 m wind speeds than coastal meteorological stations, a fact attributed to local land influences. Regarding the 10 m wind direction, models do not present large differences, although considerable deviations from recorded data were found during some dates. ARPEGE presents less scatter and lower errors than MASS when compared with QuikSCAT data.
The 10m wind fields from both atmospheric models were used to force the two selected wave models and analyse the errors and sensitivities when predicting severe wave storms. The wave model simulations show some interesting results; during the storm, the spatial wave pattern using ARPEGE showed a higher maximum, although the values of significant wave height at the buoys were lower than the ones forced by MASS (with both WAM and SWAN). The SWAN simulations show a better agreement in predicting the growing and waning of the storm peaks. The prediction of mean period was improved when using the ARPEGE wind field. However the underestimation by SWAN due to the large energy at high frequencies was evident. Validation of spectral shape predictions showed that it still has considerable error when predicting the full frequency spectra. The storms showed bimodal spectral features which were not always reproduced by wave models and are likely to be responsible for part of the discrepancies.