29, 0.11, 0.24, 0.16, 0.06, 0.16, 0.07, and 0.11 for Marseille, Barcelona, Valencia, Palma, Maó, Algiers, Offshore N, and Offshore S, respectively, which are within the range of values λλ estimated for this region by Wang et al. (2012). At the locations with lower λλ values, the PSS of Setting 8 tends to be closer to that of Setting 7, which is reasonable find protocol since Box–Cox transformation with λ=0λ=0 is log transformation. In terms of PSS, Setting 8 seems to be the best option for offshore deep water locations, but it is not clearly better for coastal nodes. Setting 8 substantially over-predicts extreme waves, showing larger positive RE values associated with the 99th HsHs percentile at the Northern Catalan coast (see Fig.
14). The above results of model performance evaluation suggest that the model Setting 5 is the best option for Catalan coast. Thus, we will use it to make projections of future wave climate in the next section. The calibrated statistical model is then applied to obtain HsHs that correspond to each of the 5 simulated SLP datasets described in Section 3.2. To diminish biases in the climate model simulations, the simulated SLP fields, denoted as Ps(t,m)Ps(t,m), are adjusted as follows: equation(24) selleckchem Pa(t,m)=σr(m)σs(m)Ps(t,m)-Ps‾(m)+Pr‾(m),where superscript r denotes the reference climate (i.e., obtained from the HIPOCAS data in this study), and X‾, the climatological mean (over the baseline
period 1971–2000) of variable X . The σs(m)σs(m) and σr(m)σr(m) are the standard deviation field of Ps(m)Ps(m) and Pr(m)Pr(m), respectively. Thus, Pa(t,m)Pa(t,m) are the simulated SLP fields that have been adjusted to have the observed baseline climate Pr(m)Pr(m) and variation scale σr(m)σr(m). The above adjustments are performed for each Thalidomide of the 5 sets of SLP simulations.
These adjusted SLP fields, Pa(t,m)Pa(t,m), are then used to derive the predictors, including P(t,m),G(t,m), and their PCs and anomalies (see Section 4 for the details). These predictors are then fed into the calibrated statistical model to obtain the corresponding HsHs. To investigate how these adjustments affect the estimated changes in HsHs between future and present, and to show the actual model biases and inter-model variability, simulations of HsHs without these adjustments are also conducted and compared with those obtained with the adjustments. Despite the shortcoming of having a few values H^s<0, Setting 5 is selected to make HsHs projections because it presents the best skill for the Catalan coast area, the focus of this study. Firstly, these projections are carried out with the predictors derived from the unadjusted model data. Their biases are assessed by comparing the projected HsHs for the present-day (baseline period) climate with the corresponding value of the HIPOCAS data (see Fig. 15). Secondly, the predictors derived from the adjusted model data are also used to obtain the HsHs projections, which are then used to assess uncertainty in wave projections.