A critical approach to clustering precipitation series in the Dobrogea region, Romania

Abstract

Listen

Translate article

This study provides a detailed framework for applying clustering algorithms to analyze precipitation data from the Dobrogea region in Romania, covering 46 meteorological stations from 1965 to 2005. Three clustering methods—K-means, K-medoids, and DBSCAN—were employed to partition the stations based on their monthly precipitation patterns. The primary goal was to outline the implementation process, highlight the use of specific R packages, and demonstrate parameter tuning to optimize clustering performance. Validation measures, including internal and stability indices, were used to assess the quality of each clustering method. While initial results indicated that K-medoids offer stable clusters and DBSCAN effectively handles noise, further comparative analysis with additional methods is necessary to determine the most suitable clustering technique for precipitation data. This work serves as a practical guide for selecting, implementing, and validating clustering algorithms in environmental data analysis.