Segmentation of Marine Forecast Zone in Korea by Using Wave Data: Based on K-Means Clustering Algorithm

The Korea Meteorological Administration (KMA) classified the forecast zone based on the administrative districts, weather characteristics, civil complaints, and location of observation points. A cluster analysis was performed using the wave observation and model data to check the validity of the current forecast zone for the wave forecast. The cluster analysis is a method used to group areas with common characteristics. In this study, a set of n objects was decomposed into k clusters (groups) and a distance-based method, k-means clustering algorithm, was used. As the data structure of the cluster analysis, the mean and standard deviation of the significant wave height and wave period during the period of 2014 to 2016 were selected. When performing the cluster analysis with wave heights and wave periods observed at ocean buoys, 11-12 cluster conditions have relatively significant internal evaluation indices suggesting the distance and variance ratio between points and points within each cluster or between distinct clusters. However, since there were few points to perform the segmentation of the marine forecast zone using only the observation data, there was a problem that clusters concentrated in a specific area were distributed. Therefore, it is necessary to perform the cluster analysis by using the KMA wave model for the segmentation of the marine forecast zone. As a result, 10 cluster condition showed the best internal evaluation indices. Although the results of the cluster analysis did not differ significantly from the current forecast zone, the segmentation was made more distinct and detailed.