This paper analyses a methodology to identify sub-series of waves parameters and wind speed able to explain the overall variability of the input dataset. To this end, the K-means clustering technique is applied to a 40-year long time series of hindcast data off the Genoa coastline (NW Italy). K-means aims to group the data in a reduced number of clusters, represented by as many modes of variability (or “model scenarios”). This work reviews and discusses a methodology to select time-varying model scenarios and assess the performances of K-means according to two indexes and increasing number of clusters. These indexes are used to compute the number of clusters best suited for the application at hand, testing different conditions as concerns the variables involved in the analysis and their temporal resolution. Results show that the indexes may not be consistent with each other, and that the number of scenarios to be reasonably employed strongly depends on how data are initially assembled. Finally, some of the model scenarios selected in front of Genoa are analysed and discussed in the framework of the local wind-wave climatology.
- Ligurian coastline
- Met-ocean modelling
ASJC Scopus subject areas
- Computer Science (miscellaneous)
- Geotechnical Engineering and Engineering Geology
- Atmospheric Science