Explainability, uncertainty & physics★ Highlight
Robustness of AI-based weather forecasts in a changing climate
Why it matters. Shows AI weather models forecast skillfully in pre-industrial and warmer climates but drift toward present-day temperatures, which bears directly on using them for climate work.
In one sentence. Shows weather-trained ML models forecast skillfully across pre-industrial, present-day and +2.9K climates, though two models show global-mean drift biases toward their training climate.
Abstract
Data-driven machine learning models for weather forecasting have made transformational progress in the last 1-2 years, with state-of-the-art ones now outperforming the best physics-based models for a wide range of skill scores. Given the strong links between weather and climate modelling, this raises the question whether machine learning models could also revolutionize climate science, for example by informing mitigation and adaptation to climate change or to generate larger ensembles for more robust uncertainty estimates. Here, we show that current state-of-the-art machine learning models trained for weather forecasting in present-day climate produce skillful forecasts across different climate states corresponding to pre-industrial, present-day, and future 2.9K warmer climates. This indicates that the dynamics shaping the weather on short timescales may not differ fundamentally in a changing climate. It also demonstrates out-of-distribution generalization capabilities of the machine learning models that are a critical prerequisite for climate applications. Nonetheless, two of the models show a global-mean cold bias in the forecasts for the future warmer climate state, i.e. they drift towards the colder present-day climate they have been trained for. A similar result is obtained for the pre-industrial case where two out of three models show a warming. We discuss possible remedies for these biases and analyze their spatial distribution, revealing complex warming and cooling patterns that are partly related to missing ocean-sea ice and land surface information in the training data. Despite these current limitations, our results suggest that data-driven machine learning models will provide powerful tools for climate science and transform established approaches by complementing conventional physics-based models.
Labels
- Topic
- Explainability, uncertainty & physics
- Type
- Evaluation or intercomparison of existing models
- Time scale
- 3–15 days, Climate projections / centennial
- Domain
- Global
- Data
- CMIP / climate model output, ERA5 / reanalysis
- Status
- Research
- Identifiers
- arXiv:2409.18529 OpenAlex:W4403798800
- Added
- 2026-10-22
Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.