Olaniyan et al. (2025) Performance Evaluation of Real-Time Sub-to-Seasonal (S2S) Rainfall Forecasts over West Africa of 2020 and 2021 Monsoon Seasons for Operational Use

Identification

• Journal: Atmosphere
• Year: 2025
• Date: 2025-09-11
• Authors: Eniola Olaniyan, Steven J. Woolnough, Felipe M. de Andrade, Linda Hirons, Elisabeth Morgan Thompson, Kamoru A. Lawal
• DOI: 10.3390/atmos16091072

Research Groups

• Department of Meteorology, African Aviation and Aerospace University, Abuja, Nigeria
• National Centre for Atmospheric Science, University of Reading, Reading, UK
• Empresa de Pesquisa Agropecuária e Extensão Rural de Santa Catarina (Epagri), Florianópolis, SC, Brazil
• Met Office, Exeter, UK
• African Centre of Meteorological Applications for Development (ACMAD), Niamey, Niger
• Climate and Development Initiative (ACDI), Department of Environmental and Geographical Science, University of Cape Town, Cape Town, South Africa

Short Summary

This study evaluates real-time ECMWF S2S rainfall forecasts during the 2020–2021 West African monsoon seasons for operational use, comparing them against satellite observations and hindcasts. The results demonstrate that ECMWF rainfall forecasts are skillful and actionable, especially up to 2–3 dekads ahead, providing confidence for early-warning and planning systems in the region.

Objective

• To evaluate the capability of operational Sub-seasonal-to-Seasonal (S2S) prediction systems over West Africa during the 2020 and 2021 monsoon seasons.
• To understand the model’s representation of monsoon dynamics and assess the accuracy of real-time S2S forecasts in capturing West African monsoon evolution.
• To quantify the influence of observational uncertainty on forecasting skill assessment by using multiple satellite datasets.
• To compare the performance of real-time forecasts against hindcast climatology to establish dependability for operational implementation.

Study Configuration

• Spatial Scale: West Africa, with a focus on four ecological zones: Gulf of Guinea (4°N–8°N), Guinea Forest (8°N–10°N), Savannah (10°N–12°N), and Sahel (12°N–16°N). All datasets were re-gridded to a 1° × 1° resolution.
• Temporal Scale: Real-time forecasts for the 2020 and 2021 West African monsoon seasons (March–October). Hindcast data from 2001–2019 for the same months. Forecast lead times were evaluated at 1–4 dekads (10-day periods) ahead, with forecasts initialized every 14 days.

Methodology and Data

• Models used: ECMWF-S2S (European Centre for Medium-Range Weather Forecasts Sub-seasonal-to-Seasonal) model, providing both real-time ensemble rainfall forecasts (51 members, 150 km resolution) and hindcasts (11 members, 150 km resolution).
• Data sources:
  • Satellite observations: Tropical Applications of Meteorology using SATellite (TAMSAT) version 3 daily rainfall estimates (5 km resolution) and Integrated Multi-satellitE Retrievals for the Global Precipitation Measurement mission (GPM-IMERG) version 6 daily rainfall data (10 km resolution, final run).
  • ECMWF-S2S database (real-time and hindcast data).
• Verification Metrics: Deterministic (bias/mean error, Pearson correlation coefficient) and probabilistic (Ranked Probability Skill Score (RPSS), Relative Operating Characteristic (ROC) curve with Area Under the Curve (AUC)) methods were used, aggregated to dekadal totals. Synchronization metric was used to quantify temporal alignment of rainfall anomalies.

Main Results

• The ECMWF-S2S model effectively captures the broad spatial and temporal structure of the West African Monsoon, including its bi-modal and uni-modal attributes and the monsoon jump.
• Forecast skill is generally higher over the Sahel than the Gulf of Guinea, and peaks during the main monsoon period (July–August).
• Forecasts achieve approximately 80% synchronization with observed rainfall-anomaly timing, indicating that roughly 8 out of 10 dekads correctly predict wet/dry phases.
• Probabilistic evaluation shows strong reliability, with high debiased ranked probability skill scores (RPSS) across various rainfall thresholds.
• The average ROC AUC of approximately 0.68 indicates moderate discrimination ability, which is above the no-skill level of 0.5.
• Systematic biases persist: the model tends to under-predict very low rains in the Gulf of Guinea and very high rains in the Sahel, and exhibits a dry bias relative to TAMSAT and a mixed wet/dry bias with GPM-IMERG.
• Actionable S2S information is most robust for the first 2–3 dekads, with useful signals persisting through the fourth dekad in certain cases.

Contributions

• Provides the first comprehensive real-time evaluation of ECMWF S2S rainfall forecasts over West Africa for operational use during the 2020–2021 monsoon seasons, addressing a critical gap in the literature that often relies on retrospective hindcasts.
• Quantifies observational uncertainty in African rainfall products by utilizing and comparing two independent satellite-derived rainfall datasets (TAMSAT and GPM-IMERG).
• Offers skill-informed operational guidance for regional climate services, demonstrating that ECMWF S2S forecasts are skillful and reliable (especially up to 2–3 dekads ahead) for early-warning and planning systems.
• Highlights specific regional and seasonal strengths and weaknesses of the ECMWF-S2S model, informing future model development and calibration efforts for West Africa.

Funding

• U.K. Research and Innovation through the Global Challenges Research Fund, Grant NE/P021077/1 (GCRF African SWIFT; 2017–2022).

Citation

@article{Olaniyan2025Performance,
  author = {Olaniyan, Eniola and Woolnough, Steven J. and Andrade, Felipe M. de and Hirons, Linda and Thompson, Elisabeth Morgan and Lawal, Kamoru A.},
  title = {Performance Evaluation of Real-Time Sub-to-Seasonal (S2S) Rainfall Forecasts over West Africa of 2020 and 2021 Monsoon Seasons for Operational Use},
  journal = {Atmosphere},
  year = {2025},
  doi = {10.3390/atmos16091072},
  url = {https://doi.org/10.3390/atmos16091072}
}