Abid et al. (2026) Analysis of spatiotemporal droughts using order statistics and archetype analysis of remotely sensed relative productivity index

Identification

• Journal: Journal of Hydrology Regional Studies
• Year: 2026
• Date: 2026-02-21
• Authors: Nesrine Abid, Abdel Hannachi, Zoubeïda Bargaoui
• DOI: 10.1016/j.ejrh.2026.103263

Research Groups

• Faculty of Sciences of Gafsa, University of Gafsa, Gafsa, Tunisia
• Department of Meteorology, Stockholm University, Sweden
• LR99ES19 Laboratory of Modelling in Hydraulics and Environment (LMHE), National Engineering School of Tunis (ENIT), University of Tunis El Manar, Tunis, Tunisia

Short Summary

This study evaluates the effectiveness of a satellite-derived productivity index (KV) combined with order statistics and archetype analysis for spatiotemporal agricultural drought monitoring in northern Tunisia, demonstrating archetype analysis as a robust method for identifying drought years and severity consistent with official reports.

Objective

• To identify and evaluate agricultural droughts in northern Tunisia using a satellite-derived productivity index (KV).
• To compare the performance of three drought identification methods: order statistics (M1), a four-class percentile-based classification (M2), and archetype analysis (M3).
• To validate the methods against official national drought reports (JORT) and assess their coherence with reported crop damage percentages.

Study Configuration

• Spatial Scale: Zaghouan county, northern Tunisia (2838 square kilometers), comprising 48 administrative districts (Imada) with areas ranging from 1 square kilometer to 161 square kilometers (median 58 square kilometers). MODIS data resolution: 500 meters by 500 meters.
• Temporal Scale: November 2000 to May 2021 (21 cereal crop seasons). MODIS data: 8-day composites, aggregated annually for the crop season.

Methodology and Data

• Models used:
  • Productivity Index (KV): Ratio of actual evapotranspiration (AET) to potential evapotranspiration (PET), also known as the Evaporative Stress Index (ESI).
  • Method M1 (Order Statistics): Drought identification based on percentiles of ranked local observations (25th percentile of the minimum, minimum of the median, 25th percentile of the median).
  • Method M2 (Percentile Classification): Four-class classification (severe, moderate, mild humid, humid) using quantiles (0.1, 0.25, 0.5, 0.75) and distance criteria (sum of absolute deviations or min-max strategy).
  • Method M3 (Archetypal Analysis - AA): A pattern recognition method that identifies extreme states (archetypes) on the convex hull of the data, representing system states as convex combinations of these archetypes.
• Data sources:
  • Satellite: MODIS (MOD16A2) for AET and PET estimates (8-day, 500 meters resolution).
  • Observation/Ground-truth: National Reports (JORT) on cereal crop damage and official drought declarations for 21 years (2000/01–2020/21) at the Imada (administrative district) level.
  • Geographic data: Digital Elevation Model (30 meters resolution), Emberger bioclimatic classification for Zaghouan county.

Main Results

• The satellite-derived productivity index (KV) proved effective for drought detection.
• Method M1a and M1b missed one drought year (2019–2020), while M1c produced one false detection (2018–2019).
• Method M2 correctly identified the four most severe drought years and four moderate drought years, aligning with JORT reports, but misclassified three JORT-declared drought years (2016–2017, 2019–2020, 2020–2021) as "mild humid."
• Archetypal analysis (M3) with three archetypes was the most accurate, correctly identifying all 11 JORT-declared drought years when using a threshold of <0.045 for the A1 (favorable crop) archetype's weight.
• With four archetypes, M3 resulted in one false alarm and one undetected year when combining A2 and A3 thresholds, or two false detections and one undetected year when using the A1 threshold.
• Archetype A1 consistently represented favorable crop conditions, while A2 and A3 represented unfavorable (drought) conditions, with A2 showing the most marked negative anomalies corresponding to the most critical droughts.
• The weights associated with favorable (A1) and unfavorable (A2, A3) archetypes correlated strongly with reported crop damage percentages, demonstrating coherence with field data.
• Archetypal analysis provided valuable geographic details at the Imada scale, highlighting regional vulnerabilities and productivity zones.

Contributions

• Proposes and validates archetype analysis as a robust, objective, spatially explicit, and threshold-independent method for agricultural drought monitoring using satellite-derived productivity indices.
• Demonstrates the effectiveness of the productivity index (KV) for drought detection in semi-arid regions, particularly in northern Tunisia.
• Provides a methodology that aligns with national drought assessments (JORT) and crop damage reports, offering a data-driven framework for agricultural damage compensation programs and adaptive water resource management.
• Highlights the advantage of archetype analysis in identifying extreme hydro-agricultural states (favorable/unfavorable) compared to methods focusing on central tendencies (e.g., K-means clustering, Principal Component Analysis).
• Contributes to achieving Sustainable Development Goal (SDG) 13 by enhancing early warning systems and adaptive water resource management in drought-prone regions.

Funding

• PAQ-Collabora (PAR&I-Tech) project, financed by the Mondial Bank, 2020, for "Suivi/élaboration d’indices de sécheresse pour l’aide à la décision concernant la gestion de la productivité céréalière en Tunisie."
• Collaboration with Tunisia's National Agricultural Insurance (la Caisse tunisienne d’assurances mutuelles agricoles -CTAMA), which manages the Compensation Fund for Agricultural Damage Resulting from Natural Disasters (Fonds d′indemnisation des dommages agricoles liés aux calamités naturelles - FIDAC).

Citation

@article{Abid2026Analysis,
  author = {Abid, Nesrine and Hannachi, Abdel and Bargaoui, Zoubeïda},
  title = {Analysis of spatiotemporal droughts using order statistics and archetype analysis of remotely sensed relative productivity index},
  journal = {Journal of Hydrology Regional Studies},
  year = {2026},
  doi = {10.1016/j.ejrh.2026.103263},
  url = {https://doi.org/10.1016/j.ejrh.2026.103263}
}