Self-Organizing Maps approaches to analyze extremes of multivariate wave climate

Abstract

Abstract. In this paper the Self-Organizing Map (SOM) technique to assess the multivariate sea wave climate at a site is analyzed and discussed with the aim of a more complete representation which includes the most severe sea states that otherwise would be missed by the standard SOM. Indeed, it is commonly recognized, and herein confirmed, that SOM is a good regressor of a sample where the density of events is high (e.g. for low/moderate and frequent sea states), while SOM fails where the density is low (e.g. for severe and rare sea states). Therefore, we have considered a trivariate wave climate (composed by significant wave height, mean wave period, and mean wave direction) collected continuously at the Acqua Alta oceanographic tower (northern Adriatic Sea, Italy) during the period 1979–2008. Three different strategies derived by the standard SOM have been tested in order to widen the range of applicability to extreme events. The first strategy contemplates a pre-processing of the input dataset with the Maximum Dissimilarity Algorithm; the second and the third strategies focus on the post-processing of SOM outputs, resulting in a two-steps SOM, where the first step is the standard SOM applied to the original dataset, and the second step is an additional SOM on the events exceeding a threshold (either taking all the events over the threshold or only the peaks of storms). Results suggest that post-processing strategies are more effective than the pre-processing one in representing the extreme wave climate, both in the time series and probability density spaces. In addition, a complete graphical representation of the outcomes of two-steps SOM as double-sided maps is proposed.