Improving global precipitation forecasts with an AI weather model trained on satellite observations
Abstract
Precipitation forecasts shape decision-making across the global economy, particularly in sectors such as agriculture. However, unlike variables such as temperature, precipitation is highly intermittent and localized, making it difficult to forecast. While recent advances in AI weather prediction systems have enabled them to surpass physical models on globally averaged metrics, improvements in mean error rarely translate to actionable forecasts of severe flooding or dry crop fields. Furthermore, most of these models are trained and evaluated against a reanalysis data product, ERA5, which has well-known biases. Here we retrain AIFS, ECMWF's widely-used, open-source operational 0.25° probabilistic graph-transformer weather model, on satellite-based precipitation observations. Our model, Laxmi, improves global probabilistic accuracy by 19% and resolves systematic distributional biases in ERA5. Specifically, Laxmi reduces drizzle overprediction by 33% for amounts less than 3 mm per day. It also mitigates extreme rainfall underprediction, improving the global 95th percentile Brier skill score by 57%. Across a case study of 10 Indian tropical storms, Laxmi delivered the most accurate forecast of 150 mm event-total precipitation in 7 events, compared to 1 for AIFS and 2 for the leading physical model, IFS. Our results demonstrate that incorporating observation-based precipitation data directly into training can substantially improve forecasts.
Related
- Learning joint probabilistic weather forecasts from station observations alone
- Can we create a `race to the top' for weather forecasts to inform smallholder farmer decisions?
- Can We Trust Evaluations of AI Weather Forecasts for Extremes? How Evaluation Protocols Shape Model Rankings
- Stress-Testing Dynamical and Generative Downscaling Using Subseasonal Extreme Precipitation Forecasts
- Global Evaluation of AI and NWP Precipitation Forecasts During Atmospheric River Events
BibTeX
@misc{schmitt2026improving,
title = {{Improving global precipitation forecasts with an AI weather model trained on satellite observations}},
author = {Julian F. Schmitt and Bertrand Delorme and Robert C. King and Yashica Patodia and Tapio Schneider and Aditi Sheshadri and Ravi Jain},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.03210},
eprint = {2609.03210},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.03210},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Improving global precipitation forecasts with an AI weather model trained on satellite observations
AU - Julian F. Schmitt
AU - Bertrand Delorme
AU - Robert C. King
AU - Yashica Patodia
AU - Tapio Schneider
AU - Aditi Sheshadri
AU - Ravi Jain
PY - 2026
DA - 2026/09/23
PB - arXiv
DO - 10.48550/arxiv.2609.03210
UR - https://arxiv.org/abs/2609.03210
AB - Precipitation forecasts shape decision-making across the global economy, particularly in sectors such as agriculture. However, unlike variables such as temperature, precipitation is highly intermittent and localized, making it difficult to forecast. While recent advances in AI weather prediction systems have enabled them to surpass physical models on globally averaged metrics, improvements in mean error rarely translate to actionable forecasts of severe flooding or dry crop fields. Furthermore, most of these models are trained and evaluated against a reanalysis data product, ERA5, which has well-known biases. Here we retrain AIFS, ECMWF's widely-used, open-source operational 0.25° probabilistic graph-transformer weather model, on satellite-based precipitation observations. Our model, Laxmi, improves global probabilistic accuracy by 19% and resolves systematic distributional biases in ERA5. Specifically, Laxmi reduces drizzle overprediction by 33% for amounts less than 3 mm per day. It also mitigates extreme rainfall underprediction, improving the global 95th percentile Brier skill score by 57%. Across a case study of 10 Indian tropical storms, Laxmi delivered the most accurate forecast of 150 mm event-total precipitation in 7 events, compared to 1 for AIFS and 2 for the leading physical model, IFS. Our results demonstrate that incorporating observation-based precipitation data directly into training can substantially improve forecasts.
ER -