Tamoffo et al. (2026) Constraining high-emission future West African monsoon with physics-weighted deep learning ensembles

Identification

• Journal: Communications Earth & Environment
• Year: 2026
• Date: 2026-07-21
• Authors: Alain Tchio Tamoffo, Fernand L. Mouassom, Torsten Weber, Daniela Jacob
• DOI: 10.1038/s43247-026-03839-8

Research Groups

• Climate Service Center Germany (GERICS), Helmholtz-Zentrum Hereon, Hamburg, Germany.
• Department of Mathematics and Statistics, Memorial University of Newfoundland, St. John’s, Newfoundland, Canada.

Short Summary

This study proposes a physics-guided artificial neural network (ANN) approach to constrain regional climate projections by reducing the inter-model spread in West African monsoon rainfall under high-emission scenarios. The method uses sea-level pressure patterns from the Sahelian monsoon ocean-pressure index (SMOPI) as predictors and evaluates the physical realism of learned teleconnections against JRA-55 reanalysis.

Objective

• To develop a physics-guided ANN approach to constrain regional climate projections.
• To reduce the inter-model spread in West African monsoon rainfall under high-emission scenarios.

Study Configuration

• Spatial Scale: Regional (West Africa).
• Temporal Scale: Multi-decadal (1958–2100).

Methodology and Data

• Models used: Physics-guided artificial neural networks (ANNs) trained on CMIP6 models.
• Data sources: Sea-level pressure patterns from the Sahelian monsoon ocean-pressure index (SMOPI), JRA-55 reanalysis, and CMIP6 outputs.

Main Results

• The physics-guided ANN approach reduces the inter-model spread in West African monsoon rainfall by 19.7% for high-skill models.
• The method improves the cumulative distribution functions (CDFs) of precipitation, making them closer to historical observations.
• The ANNs-based projections preserve the magnitude and structure of the raw CMIP6 climate change signal.

Contributions

• This study provides a new approach to constrain regional climate projections by reducing the inter-model spread in West African monsoon rainfall under high-emission scenarios.
• The physics-guided ANN method offers a more robust scientific basis for resilience policies in West Africa.

Funding

• This research was funded by the Climate Service Center Germany (GERICS) and the Helmholtz-Zentrum Hereon.

Citation

@article{Tamoffo2026Constraining,
  author = {Tamoffo, Alain Tchio and Mouassom, Fernand L. and Weber, Torsten and Jacob, Daniela},
  title = {Constraining high-emission future West African monsoon with physics-weighted deep learning ensembles},
  journal = {Communications Earth & Environment},
  year = {2026},
  doi = {10.1038/s43247-026-03839-8},
  url = {https://doi.org/10.1038/s43247-026-03839-8}
}