Pérez et al. (2026) Quantifying the Value of Ai-Based Meteorological Postprocessing for Seasonal Hydrological Forecasting in Mediterranean Semi-Arid Basins

Identification

• Journal: SSRN Electronic Journal
• Year: 2026
• Authors: David Ricardo De León Pérez, Dariana Isamel Avila-Velasquez, Hector Macian‐Sorribes, Sergio Salazar-Galán, Manuel Pulido-Velazquez, Félix Francés
• DOI: 10.2139/ssrn.6118139

Short Summary

This study quantifies the value of AI-based fuzzy rule systems for postprocessing meteorological inputs in seasonal hydrological forecasts in Spain's Jucar River Basin, finding that forecast reliability coverage increased substantially from 41.6% to 73.7% and extending the operational lead time from 1–2 months to 4–5 months ahead.

Objective

• Quantify the impact and value of AI-based meteorological postprocessing (specifically fuzzy rule-based systems) on the performance of ensemble seasonal hydrological forecasts in Mediterranean semi-arid basins, and systematically differentiate between meteorological and hydrological uncertainties.

Study Configuration

• Spatial Scale: Jucar River Basin, Spain (Mediterranean semi-arid basin).
• Temporal Scale: 20 years (1995–2014).

Methodology and Data

• Models used: TETIS distributed hydrological model; Fuzzy rule-based systems (AI-based postprocessing for bias correction).
• Data sources: Four global seasonal forecast systems (ECMWF-SEAS5, Météo-France System8, DWD-GCFS2.1, CMCC-SPSv35); Dual evaluation frameworks (Proxy-Truth and Observed-Truth) were employed.

Main Results

• Postprocessing substantially improved forecast reliability, increasing predictive uncertainty coverage from 41.6% to 73.7%.
• Probabilistic skill (CRPSS) improved across lead times, ranging from 0.068 to 0.131.
• The operational lead time for effective water management was extended from 1–2 months to 4–5 months ahead.
• Severely deficient systems showed transformative enhancement, with uncertainty coverage increasing from 24.6% to 64.3%.
• Effectiveness exhibited strong spatial heterogeneity: mountain headwater stations demonstrated exceptional responsiveness (24.4 percentage points improvement), while valley locations showed minimal benefits (2.9 percentage points improvement).

Contributions

• Systematic quantification of the value of AI-based meteorological postprocessing for reducing systematic upstream biases in seasonal hydrological forecasts in semi-arid regions.
• Demonstration that enhanced forecasts can extend the operational lead time significantly (up to 4–5 months), providing unprecedented capability for proactive drought planning and resource allocation.
• Identification that topographic characteristics significantly modulate postprocessing effectiveness, highlighting the spatial heterogeneity of forecast improvements.

Funding

• Not specified in the provided text.

Citation

@article{Pérez2026Quantifying,
  author = {Pérez, David  Ricardo De León and Avila-Velasquez, Dariana  Isamel and Macian‐Sorribes, Hector and Salazar-Galán, Sergio and Pulido-Velazquez, Manuel and Francés, Félix},
  title = {Quantifying the Value of Ai-Based Meteorological Postprocessing for Seasonal Hydrological Forecasting in Mediterranean Semi-Arid Basins},
  journal = {SSRN Electronic Journal},
  year = {2026},
  doi = {10.2139/ssrn.6118139},
  url = {https://doi.org/10.2139/ssrn.6118139}
}

Generated by BiblioAssistant using gemini-flash-latest (Google API)