Robust Panel Clustering of Station-Scale Precipitation in Türkiye
Coğrafi Bilimler Dergisi · 30.04.2026 · 24 (1) · 324-339
DOI: 10.33688/aucbd.1695834 ↗Kurz gefasst
Fachartikel in der Zeitschrift Coğrafi Bilimler Dergisi, online erschienen am 30.04.2026. Verfasst von Çağlar Sözen und 1 weiteren Autor:innen. Die Arbeit nennt 29 Literaturangaben.
Abstract (Originalsprache)
This study uses station-level annual precipitation series for Türkiye to objectively delineate homogeneous precipitation regimes. We first applied k-means, PAM, and hierarchical (agglomerative) clustering for K=2"–" 10 and evaluated the solutions using the APN, AD, ADM, FOM, Connectivity, Dunn, and Silhouette indices. Although classical algorithms favored K=2, Rize acted as a high-leverage observation, exerting a disproportionate influence on centroid estimation, producing a singleton cluster, and obscuring meaningful substructure among the remaining stations. We therefore adopted a robust trimmed k-means approach. A multi-index assessment identified K=4 as optimal for APN/ADM/FOM, yielding four structural regimes (C1–C4) while treating Rize as a trimmed singleton. The resulting partition quantifies—at the station scale—the well-known contrast between coastal/orographic enhancement and continental drying, producing clusters that are more compact, less overlapping, and climatologically interpretable.
Zitieren
APA
Sözen, Ç., & Bulut, H. (2026). Robust Panel Clustering of Station-Scale Precipitation in Türkiye. Coğrafi Bilimler Dergisi, 24(1), 324-339. https://doi.org/10.33688/aucbd.1695834
Vancouver
Sözen Ç, Bulut H. Robust Panel Clustering of Station-Scale Precipitation in Türkiye. Coğrafi Bilimler Dergisi. 2026;24(1):324-339. doi:10.33688/aucbd.1695834
BibTeX
@article{Szen2026,
author = {Sözen, Çağlar and Bulut, Hasan},
title = {Robust Panel Clustering of Station-Scale Precipitation in Türkiye},
journal = {Coğrafi Bilimler Dergisi},
year = {2026},
volume = {24},
number = {1},
pages = {324--339},
doi = {10.33688/aucbd.1695834}
}
EUVIMED