Statistical characterization of Andalusian wave climate for several combinations of Global Climate Models and Regional Climate Models and periods 2026 - 2045 and 2081 - 2100.

The following text is an extract of the extended abstract entitled "Parametric Characterization of Wave Climate along the Andalusian Coast for Non-Stationary Stochastic Simulation" whose authors are Manuel Cobos, Pedro Magaña, Pedro Otiñar and Asunción Baquerizo, and that was included in proceedings of 39th IAHR World Congress where this dataset is included.

Processed data comes from PIMA Adapta Costas project (Ramírez et al., 2019), in particular, from projections of maritime climate for 2026-2045 and 2081-2100. Sea climate contains, among other information, time series of the significant wave height (H_s) obtained for several combinations of GCM-RCM projections of EUR-11 for the RCP 8.5. GCM-RCM combinations ACCE, CMCC, CNRM, GFDL, HADG, IPSL, MIRO with a 0.1 degrees grid were used for the Atlantic facade while CNRM, HADG, IPSL, MIRO, MEDC, MPIE, ESM2, EART models with 1/11 degrees were used for the Mediterranean one. A total of 210 locations were analyzed, 54 at the Atlantic facade and 156 at the Mediterranean one (Figure 1). The data was bias adjusted using the Empirical Quantile Mapping (Déqué et al., 2007; Michelangeli et al., 2009). Information of the significant wave height and the dependence between the values at a given time with previous values with a VAR(q) model is already available.

At each location, the methodology of Lira-Loarca et al. (2021) was applied, using the software described in Cobos et al. (2022a). More precisely, for every GCM-RCM (hereinafter, model n for n = 1, .., N where N = 7 for Atlantic data and N = 8 for the Mediterranean data), a non-stationary marginal distribution of H_s, , assuming that the year was the largest periodicity of the climate, was fitted to data using a lognormal model for the central part and two generalized Pareto distribution for the lower and upper tails, as in Solari and Losada (2011). The non- stationarity is considered by assuming a decomposition of the parameters of the distribution and of the percentiles of the common end points of the interval into a trigonometric truncated expansion.

In addition, the coefficients of the matrix, C_n, of a VAR(q) model with q up to 92 hours were estimated. The ensemble multi-model characteristics of the data were obtained from the compound distributions and the weighted averaged matrix coefficients.

Soon, the results of the peak period (T_p) and mean incoming wave direction (ϑ_m) and the coefficients of the multivariate VAR model will also be included.