EVALUATION OF EXTREME SEA STATES BY INTEGRATING EXPERIMENTAL DATA WITH THE RESULTS OF WEATHER AND WAVE MODELS

An essential tool of engineering is the estimation of the probability of of natural events, such as rainfall, river flow or - in the case of offshore or coastal structures – the sea state. This estimate is normally carried out by adapt-ing suitable theoretical frequency distributions to the extreme values to the data series. In the case of marine struc-tures, the most important parameter is the significant wave height (SWH), whose values are generally provided by wave meters in situ (buoys, or sometimes pressure gauges, pole, or platform gauges) and more recently by satellite altimetry data. There are however relatively few sites where wave meters are present, and even fewer have been kept and monitored long enough to provide reliable time series. Over the past 25 years, many state and international meteorological centres, as well as some research insti-tutes and private companies have begun to systematically run global and regional wave generation and propaga-tion models. These models are in turn driven by meteorological forecasting systems and constantly validated through the acquisition ('assimilation') of measured data. Both forecast and analysis are published almost in real time, thus providing long time series of "synthetic data" which are today the main source of information for statis-tical analyses, to the point that the use of such data has now become a commonplace. However, the estimation of extreme SWH values for high return times through synthetic data raises several questions: apart from the obvious problem of reliability of the modelling chains, an important aspect is the way through which the wave data measured in situ are assimilated into the analysis. Most of the assimilation procedures are based on satellite alti-metric data, which are scattered in time (at intervals of many hours) and distant in space (tens or hundreds of kil-ometers); the extreme SWH values are thus often missed. Furthermore, the sampling time of the models, i.e. the time interval in which the data are stored and released, is often longer than the standard sampling time of the buoys, which causes a negative distortion of the estimated extreme values (Arena et al., 2013; Reale et al., 2014; Dentale et al. 2016). In order to overcome these problems, a procedure has been proposed by some of the Authors of this work (Dentale et al 2018, 2020) according to which the extreme SWH curves as a function of the return time TR (herein-after: SWH (TR)) resulting from the synthetic data are compared and calibrated with similar curves calculated from the buoy data. This document presents an extension of this technique.