Zarei (2026) Medium -term monitoring and machine learning-based forecasting of drought dynamics in Iran

Identification

• Journal: Scientific Reports
• Year: 2026
• Date: 2026-03-20
• Authors: Abdol Rassoul Zarei
• DOI: 10.1038/s41598-026-45031-0

Research Groups

• Department of Range and Watershed Management (Nature engineering), College of Agriculture, Fasa University, Fasa, Iran

Short Summary

This study comprehensively assesses historical drought conditions in Iran from 1967 to 2024 and forecasts decadal drought dynamics for 2025–2036 using climate observations and machine learning, revealing a projected significant increase in drier conditions and the disappearance of extreme wet periods.

Objective

• To comprehensively assess past drought conditions and analyze their trends in Iran using climate observations.
• To develop and validate a robust machine learning-based model for decadal drought forecasting in Iran.
• To forecast future drought conditions (2025–2036) and identify projected shifts in drought severity and frequency across the country.

Study Configuration

• Spatial Scale: Iran, utilizing data from 34 meteorological stations.
• Temporal Scale: Historical analysis spanning 58 years (1967–2024); decadal forecast for 2025–2036. Drought conditions were assessed at 1-month, 3-month, and 12-month time scales.

Methodology and Data

• Models used:
  • Drought Index: Reconnaissance Drought Index (RDI)
  • Optimization Algorithms: Pelican Optimization Algorithm (POA), Genetic Algorithm (GA), Particle Swarm Optimization (PSO)
  • Machine Learning Models: Support Vector Machine (SVM), Bidirectional Long Short-Term Memory (BiD-LSTM)
  • Selected Forecasting Framework: SVM-POA (Support Vector Machine optimized with Pelican Optimization Algorithm)
  • Trend Analysis: Mann-Kendall trend test (inferred from references)
  • Performance Metrics: R² (coefficient of determination), Mean Absolute Error (MAE), Normalized Root Mean Squared Error (NRMSE), Kappa index.
• Data sources:
  • 58 years (1967–2024) of climate observations from 34 meteorological stations across Iran.
  • Data provided by the Iranian Meteorological Organization.

Main Results

• The Pelican Optimization Algorithm (POA) demonstrated superior performance in optimizing the Support Vector Machine (SVM) compared to Genetic Algorithm (GA) and Particle Swarm Optimization (PSO).
• The SVM-POA model was identified as the most accurate forecasting model, outperforming the Bidirectional Long Short-Term Memory (BiD-LSTM) network based on multiple performance metrics.
• Historically, the "Normal" (N) RDI class was the most frequent drought condition in Iran.
• Decadal forecasts for 2025–2036 indicate that "Normal" (N) and "Moderately Dry" (M-D) classes will become the most frequent, with extreme wet conditions expected to disappear.
• Trend diagnostics revealed a prevailing shift toward drier conditions across all 34 meteorological stations, regardless of statistical significance.
• The proportion of stations exhibiting statistically significant drying trends is projected to increase sharply between the historical and extended periods:
  • For the 1-month scale: from 26.47% to 88.23%.
  • For the 3-month scale: from 2.9% to 35.29%.
  • For the 12-month scale: the proportion of stations with significant drying trends remains unchanged.
• These findings underscore a high likelihood of increasingly severe and frequent droughts in Iran in the coming decade.

Contributions

• Provides a comprehensive historical assessment and decadal forecast of drought dynamics specifically for Iran, a region highly vulnerable to water scarcity.
• Validates and applies an optimized machine learning framework (SVM-POA) for medium-term drought forecasting, demonstrating its superior accuracy compared to other benchmark models.
• Quantifies the projected increase in the prevalence and severity of drier conditions across Iran, offering critical foresight into future water resource challenges.
• Offers crucial scientific evidence to support the development of urgent, forward-looking strategies for sustainable water resource management and drought mitigation in Iran.

Funding

• Not explicitly stated in the provided text, beyond acknowledging the Iran Meteorological Organization for data provision.

Citation

@article{Zarei2026Medium,
  author = {Zarei, Abdol Rassoul},
  title = {Medium -term monitoring and machine learning-based forecasting of drought dynamics in Iran},
  journal = {Scientific Reports},
  year = {2026},
  doi = {10.1038/s41598-026-45031-0},
  url = {https://doi.org/10.1038/s41598-026-45031-0}
}