After the simulated SLP data being adjusted to have the observed baseline climate and variation scale, the bias for the present-day median HsHs (see Fig. 17) almost disappears completely, as would be expected. The adjustments also affect the projected changes in HsHs; they attenuate the projected relative changes in general (especially for models driven by ECHAM5), although the pattern of change is maintained. It is not possible to know which projected changes are more reliable, because any type of statistical adjustments has its own limitations. In particular, such adjustments selleck compound cannot account

for any feedback (e.g., how changes in ocean waves may affect changes in SLP) that may exist in the real world. Similarly, Fig. 18 and Fig. 19 show the present-day climate and projected changes of the 50-year return value of HsHs (z50z50). The model bias patterns (compare upper panels of Fig. 18 with right panel of Fig.

15) are similar to those for the median HsHs, showing in general significant HIR_E overestimation and moderate or low overestimation by the other models. The projected future changes (Fig. 18, lower panels) vary more between models than for Cyclopamine mouse the median HsHs, as similarly found by Casas-Prat and Sierra (2013). These results are reasonable because extreme values are normally exposed to a larger uncertainty. Along the Catalan coast, there is a general tendency for z50z50 to decrease or remain constant, except in the northern coast where models RCA_E and HIR_E project an increase. The maximum rate of change

is around Fludarabine mw ±20%±20% (larger than for the median HsHs) which is in agreement with the non-linear relation between HsHs and wind for wind-sea states, typically present in stormy conditions, as pointed out by Casas-Prat and Sierra (2013). Very similar spatial patterns and magnitudes of change were obtained by Casas-Prat and Sierra (2013) for the models REM_E and RCA_E. On the contrary, the projected change that they obtained for RCA_H differed from the present study, obtaining a notable increase of z50z50 along almost all E-facing coasts. The adjustments to the simulated SLP data reduce the current z50z50 but not necessarily the model bias. For example, among the five sets of RCM–GCM simulations, HIR_E has the largest positive bias before the adjustments, but it has a negative bias after the adjustments. As for the median HsHs, after applying the adjustments (Fig. 19), the magnitude of change in the z50z50 is slightly reduced, but to much lesser extent than for the median HsHs. Indeed, the projected changes of z50z50 are barely the same (compare Fig. 18 and Fig. 19). This study proposes a statistical method to model near-shore HsHs, at a 3 h and 25 km resolution. This high spatial–temporal resolution is suitable for coastal impact analysis although a complete assessment would have to involve additional wave parameters, such as wave direction (Reguero et al., 2011).