Tofiq et al. (2026) Analyzing Drought Vulnerability with Clustering: A Study of Southeast Türkiye Using Multiple Drought Indices

Identification

• Journal: Pure and Applied Geophysics
• Year: 2026
• Date: 2026-01-28
• Authors: Fahid Abbas Tofiq, Nermin Şarlak
• DOI: 10.1007/s00024-026-03914-3

Research Groups

• Department of Civil Engineering, Erbil Technical Engineering College, Erbil Polytechnic University, Erbil, Iraq
• Department of Civil Engineering, Konya Technical University, Konya, Tu¨rkiye

Short Summary

This study investigates drought patterns in Southeast Tu¨rkiye by clustering regional drought behaviour using four widely applied indices (SPI, SPEI, PDSI, scPDSI) and hierarchical clustering. It identifies two primary clusters, with high-altitude, snow-covered regions showing steeper drying trends and increased vulnerability due to climate change.

Objective

• To investigate drought patterns in Southeast Tu¨rkiye by clustering regional drought behaviour using multiple drought indices.
• To identify spatially homogeneous drought-prone areas to optimize water management, agricultural planning, and targeted mitigation strategies.
• To evaluate the performance of Dynamic Time Warping (DTW) and Euclidean Distance (ED) as distance metrics for clustering drought indices.

Study Configuration

• Spatial Scale: Southeast Tu¨rkiye and a part of Eastern Anatolia Region, covering 29 meteorological stations. Elevations range from 347 meters to 2,286 meters above sea level.
• Temporal Scale: Monthly data from January 1965 to December 2021 (57 years).

Methodology and Data

• Models used:
  • Drought Indices: Standardized Precipitation Index (SPI-12), Standardized Precipitation Evapotranspiration Index (SPEI-12), Palmer Drought Severity Index (PDSI), self-calibrated PDSI (scPDSI).
  • Clustering: Hierarchical Clustering (HC) with Dynamic Time Warping (DTW) and Euclidean Distance (ED) as distance metrics.
  • Optimal Cluster Number: Silhouette method, Elbow method.
  • Cluster Validation: Davies–Bouldin Index (DBI), Calinski–Harabasz Index (CHI).
  • Trend Analysis: Sen’s slope test, Mann–Kendall test.
  • Potential Evapotranspiration (PET) estimation: Thornthwaite method.
  • Missing Data Filling: Simple arithmetic method, correlation with nearby stations.
  • Homogeneity Tests: Standard Normal Homogeneity Test (SNHT), Pettitt test, Buishand test.
  • Outlier Detection: Interquartile Range (IQR), Z-score tests.
  • Contingency Analysis: Hits (H), Misses (M), False Alarms (FA), Correct Wet (CW), False Alarm Rate (FAR), Hits Rate (HR), Bias (FBI), Proportion Correct (PC).
• Data sources: Monthly total precipitation and average temperature data from 29 meteorological stations, collected from the General Directorate of Meteorology / Ministry of Environment, Urbanization and Climate Change, Tu¨rkiye.

Main Results

• Hierarchical clustering, optimized using Silhouette and Elbow methods, consistently identified two primary clusters across 29 stations for all four drought indices (SPI-12, SPEI-12, PDSI, scPDSI).
• Cluster validation showed that Dynamic Time Warping (DTW) generally outperformed Euclidean Distance (ED) for indices incorporating precipitation and evapotranspiration dynamics (SPEI-12, scPDSI, and SPI-12 Cluster 1), yielding lower mean Davies–Bouldin Index (DBI) values (e.g., SPEI-12: 1.04 vs. 1.08; scPDSI: 1.20 vs. 1.35) and higher Calinski–Harabasz Index (CHI) scores in five of eight clusters.
• ED performed better for PDSI (mean DBI = 0.99 vs. 1.21) and SPI-12 Cluster 2.
• Sen’s slope and Mann–Kendall analyses revealed significant drying trends across all indices and clusters in Southeast Tu¨rkiye.
• Cluster 2, largely representing high-altitude, snow-covered regions, consistently showed steeper drying trends for SPI-12, SPEI-12, and scPDSI, indicating increased vulnerability due to declining snowpack and shifts in spring season.
• Contingency analysis indicated that SPEI-12 effectively captured dry and wet periods similar to SPI-12, while scPDSI showed only marginal improvement over PDSI in drought classification consistency, suggesting that region-specific adaptations did not have the expected effect.

Contributions

• Provides a comprehensive regionalization of drought patterns in Southeast Tu¨rkiye using multiple drought indices and advanced clustering techniques (DTW and ED).
• Systematically evaluates the performance of DTW and ED distance metrics for clustering drought indices, offering insights into their suitability based on index characteristics.
• Identifies specific high-altitude, snow-covered regions (Cluster 2) as particularly vulnerable to increasing drought severity due to climate change impacts on snowpack and water availability.
• Emphasizes the critical need for region-specific drought management strategies and adjustments to drought index inputs (e.g., Available Water Capacity for snow-affected areas) for improved reliability.

Funding

• Open access funding provided by the Scientific and Technological Research Council of Tu¨rkiye (TU¨ BI˙TAK).
• No other funds, grants, or support were received during the preparation of this manuscript.

Citation

@article{Tofiq2026Analyzing,
  author = {Tofiq, Fahid Abbas and Şarlak, Nermin},
  title = {Analyzing Drought Vulnerability with Clustering: A Study of Southeast Türkiye Using Multiple Drought Indices},
  journal = {Pure and Applied Geophysics},
  year = {2026},
  doi = {10.1007/s00024-026-03914-3},
  url = {https://doi.org/10.1007/s00024-026-03914-3}
}