Avila-Velasquez et al. (2025) How Good Are Drought Forecasts? Skill of multi-model Seasonal Forecast of Meteorological Droughts in a semi-arid Mediterranean Basin

Identification

• Journal: Earth Systems and Environment
• Year: 2025
• Date: 2025-12-12
• Authors: Dariana Isamel Avila-Velasquez, Hector Macian-Sorribes, Manuel Pulido-Velazquez
• DOI: 10.1007/s41748-025-00965-9

Research Groups

• Institute of Water and Environmental Engineering (IIAMA), Universitat Politècnica de València, Valencia, Spain

Short Summary

This study develops and evaluates a multi-model seasonal forecasting system for meteorological drought indices (SPI and SPEI) in the semi-arid Jucar River Basin, Spain, integrating Copernicus Climate Change Service (C3S) forecasts with artificial intelligence post-processing to demonstrate high forecast skill for water management.

Objective

• To develop and evaluate a multi-model seasonal forecasting system for meteorological drought indices (Standardized Precipitation Index - SPI and Standardized Precipitation-Evapotranspiration Index - SPEI) at multiple aggregation scales.
• To assess the forecast skill of this system using hindcasts from 1995–2014 in the semi-arid Jucar River Basin, Spain, and demonstrate its applicability for drought early warning and water management.

Study Configuration

• Spatial Scale: Jucar River Basin District (JRBD), Eastern Iberian Peninsula, Spain, with a total surface area of 42,756 square kilometers. Data processed on a 0.25° x 0.25° grid resolution.
• Temporal Scale:
  • Hindcast evaluation period: 1995–2014.
  • Reference period for drought index calculation: 1973–2022 (50 years).
  • Forecast horizon: Up to 6 lead months.
  • Drought index aggregation scales: 6, 12, 18, and 24 months.

Methodology and Data

• Models used:
  • ECMWF-SEAS5 (European Centre for Medium-Range Weather Forecasts Seasonal Forecasting System 5)
  • Météo-France System8
  • DWD-GCFS2.1 (Deutscher Wetterdienst Global Climate Forecast System 2.1)
  • CMCC-SPSv3.5 (Centro Euro-Mediterraneo sui Cambiamenti Climatici Seasonal Prediction System v3.5)
  • Multi-Model Ensemble (MME) combining the four individual systems.
• Data sources:
  • Seasonal forecasts (hindcasts) from the Copernicus Climate Change Service (C3S) with a native spatial resolution of 1° x 1°.
  • ERA5 reanalysis product (precipitation, mean, minimum, and maximum temperature) from C3S for reference observations and post-processing, with a 0.25° x 0.25° resolution.
  • Artificial Intelligence (Sugeno fuzzy rule-based systems of order 1) for post-processing (bias correction and downscaling) of raw seasonal forecasts against ERA5.
  • Drought indices: Standardized Precipitation Index (SPI) and Standardized Precipitation-Evapotranspiration Index (SPEI), calculated using the gamma distribution.
  • Reference evapotranspiration (ET0) calculated using the Hargreaves method.
  • Forecast skill evaluated using the Continuous Ranked Probability Skill Score (CRPSS).

Main Results

• The forecasting system shows high skill, with values around 90% at one lead month.
• Skill remains above 64% for SPI-6 and 67% for SPEI-6 at three lead months.
• Longer aggregation scales (12, 18, and 24 months) retain useful skill up to five lead months, and for the entire 6-month forecasting horizon.
• The Multi-Model Ensemble (MME) consistently outperforms individual forecasting systems, showing a global improvement of approximately 1–2% in skill across all lead months, particularly for the SPEI index.
• SPEI generally demonstrates higher skill percentages and less variability compared to SPI, especially in winter and summer, attributed to the inclusion of temperature.
• Seasonal variations in skill are observed: ECMWF-SEAS5 excels at short lead times (0-2 months), CMCC-SPSv3.5 at intermediate lead months (3-5 months), and Météo-France System8 at longer lead times (6 months) during winter and summer.
• Spatially, coastal areas near the Mediterranean Sea show lower skill for SPI-6, while inland continental regions exhibit more stable and higher predictive reliability. For aggregations of 12 months and longer, very high skill (>80%) dominates most of the basin.

Contributions

• Development of an innovative multi-model seasonal forecasting system for SPI and SPEI, integrating C3S forecasts with artificial intelligence (fuzzy logic) post-processing, enhancing robustness and reducing uncertainty.
• Comprehensive evaluation of meteorological drought index forecast skill across multiple temporal aggregations (6, 12, 18, 24 months) and lead times (up to 6 months) in a semi-arid Mediterranean basin.
• Demonstration of the added value of a multi-model ensemble approach and the complementary use of SPI and SPEI for a more robust and comprehensive view of drought conditions.
• Highlights the superior performance of fuzzy logic for bias correction and downscaling of seasonal climate forecasts compared to standard methods.
• Proposes a transferable methodological framework for drought early warning and water management, supported by an operational web-based implementation, applicable to other drought-prone regions.

Funding

• University Teacher Training (FPU) contract of the Ministry of Universities (FPU20/0749).
• Project “INtegrated FORecasting System for Water and the Environment (WATER4CAST)” (ref: PROMETEO/2021/074) from the Program for the promotion of scientific research, technological development and innovation in the Valencian Community for research groups of excellence, Department of Innovation, Universities, Science and Digital Society, Generalitat Valenciana.

Citation

@article{AvilaVelasquez2025How,
  author = {Avila-Velasquez, Dariana Isamel and Macian-Sorribes, Hector and Pulido-Velazquez, Manuel},
  title = {How Good Are Drought Forecasts? Skill of multi-model Seasonal Forecast of Meteorological Droughts in a semi-arid Mediterranean Basin},
  journal = {Earth Systems and Environment},
  year = {2025},
  doi = {10.1007/s41748-025-00965-9},
  url = {https://doi.org/10.1007/s41748-025-00965-9}
}