Suitable Measures for the Potential Operational Utility of AI NWP Rainfall Forecasts Over Africa
Abstract
Artificial intelligence (AI)-based weather prediction is approaching the skill of physical numerical weather prediction (NWP) systems at a fraction of the computational cost. This is particularly promising for Africa, where rainfall extremes are intensifying and many forecasting centres lack the infrastructure to run physical models at extended lead times. We present a calibrated comparison of GraphCast, GenCast and the Functional Generative Network (FGN) against the physical NWP model IFS for rainfall prediction across Africa. Deterministic and probabilistic forecasts are postprocessed using Isotonic Distributional Regression and evaluated with the Continuous Ranked Probability Score against IMERG, RFEv2 and CHIRPS across seasons, wet and dry regimes, elevation zones and lead times. All models retain skill beyond climatology across most seasons and at extended lead times. AI models generally outperform IFS in wet regions, whereas IFS performs better in dry, high-elevation areas, where its finer resolution better represents orographic controls on rainfall. Across observational datasets and seasons, AI models achieve a median improvement of approximately 5% over IFS. GraphCast achieves calibrated skill comparable to the ensemble-based FGN, although FGN provides greater significant skill at longer lead times. These results highlight the potential of calibrated AI weather prediction to provide accessible and computationally efficient rainfall forecasts across Africa, while demonstrating the continuing importance of spatial resolution, ensemble design and regional characteristics.
Related
- Spatial identification of areas suitable for other effective area‐based conservation measures in the European Union
- Suitable habitats redistribution for the European pond turtle (Emys orbicularis) lineages under future climate change
- Assessment of Rainfall Variability and Its Impact on Soil Erosion Risk Using Rainfall Erosivity in Kwara State, Nigeria
- A systematic method for selecting suitable locations for an urban climatological sensor observation network
- Rainfall and salinity-related landscapes shape cholera persistence in Haiti
BibTeX
@misc{nath2026suitable,
title = {{Suitable Measures for the Potential Operational Utility of AI NWP Rainfall Forecasts Over Africa}},
author = {Shruti Nath and Docko Sow and Koomi Toussaint Amoussouvi and Fenwick Cooper and Josiah Kiarie Kimani and John Bagiliko and Florian Pappenberger and Rendani Mbuvha},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.31775},
eprint = {2609.31775},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.31775},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Suitable Measures for the Potential Operational Utility of AI NWP Rainfall Forecasts Over Africa
AU - Shruti Nath
AU - Docko Sow
AU - Koomi Toussaint Amoussouvi
AU - Fenwick Cooper
AU - Josiah Kiarie Kimani
AU - John Bagiliko
AU - Florian Pappenberger
AU - Rendani Mbuvha
PY - 2026
DA - 2026/09/24
PB - arXiv
DO - 10.48550/arxiv.2609.31775
UR - https://arxiv.org/abs/2609.31775
AB - Artificial intelligence (AI)-based weather prediction is approaching the skill of physical numerical weather prediction (NWP) systems at a fraction of the computational cost. This is particularly promising for Africa, where rainfall extremes are intensifying and many forecasting centres lack the infrastructure to run physical models at extended lead times. We present a calibrated comparison of GraphCast, GenCast and the Functional Generative Network (FGN) against the physical NWP model IFS for rainfall prediction across Africa. Deterministic and probabilistic forecasts are postprocessed using Isotonic Distributional Regression and evaluated with the Continuous Ranked Probability Score against IMERG, RFEv2 and CHIRPS across seasons, wet and dry regimes, elevation zones and lead times. All models retain skill beyond climatology across most seasons and at extended lead times. AI models generally outperform IFS in wet regions, whereas IFS performs better in dry, high-elevation areas, where its finer resolution better represents orographic controls on rainfall. Across observational datasets and seasons, AI models achieve a median improvement of approximately 5% over IFS. GraphCast achieves calibrated skill comparable to the ensemble-based FGN, although FGN provides greater significant skill at longer lead times. These results highlight the potential of calibrated AI weather prediction to provide accessible and computationally efficient rainfall forecasts across Africa, while demonstrating the continuing importance of spatial resolution, ensemble design and regional characteristics.
ER -