# Do AI weather models miss extremes?

arXiv: https://arxiv.org/abs/2608.09972 · HTML: https://mlsanity.com/paper/2608.09972/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 9 of its 30 references that are in the map.
- First posted on arXiv 31 Jul 2026.

## Builds on (most cited first, 9 of 9)

- [FourCastNet: A Global Data-driven High-resolution Weather Model using Adaptive Fourier Neural Operators](https://mlsanity.com/paper/2202.11214/) (2022)
- [Kilometer-Scale Convection Allowing Model Emulation using Generative Diffusion Modeling](https://mlsanity.com/paper/2408.10958/) (2024)
- [Generative AI for fast and accurate statistical computation of fluids](https://mlsanity.com/paper/2409.18359/) (2024)
- [Fixing the Double Penalty in Data-Driven Weather Forecasting Through a Modified Spherical Harmonic Loss Function](https://mlsanity.com/paper/2501.19374/) (2025)
- [Do AI models produce better weather forecasts than physics-based models? A quantitative evaluation case study of Storm Ciar'an](https://mlsanity.com/paper/2312.02658/) (2023)
- [WeatherReal: A Benchmark Based on In-Situ Observations for Evaluating Weather Models](https://mlsanity.com/paper/2409.09371/) (2024)
- [Universal Diffusion-Based Probabilistic Downscaling](https://mlsanity.com/paper/2602.11893/) (2026)
- [EPT-2 Technical Report](https://mlsanity.com/paper/2507.09703/) (2025)
- [MAUSAM: An Observations-focused assessment of Global AI Weather Prediction Models During the South Asian Monsoon](https://mlsanity.com/paper/2509.01879/) (2025)

## Abstract (arXiv)

AI weather models are often reported to underestimate extremes, but most evidence concerns deterministic regression models verified against reanalysis. We evaluate twelve physical and AI forecast models against ECMWF IFS using ten months of European station observations. The evaluation covers 10 m wind, 2 m temperature, solar radiation, and precipitation within regimes defined from a fixed ERA5 1991-2020 climatology. We find no uniform AI-specific deficit in the tails. Several AI models remain more accurate than IFS under extreme conditions, while others deteriorate markedly; comparable variation occurs among physical models. Every model nevertheless exhibits a common conditional-error pattern, overpredicting low observations and underpredicting high observations. Attenuation of extreme values therefore does not imply a uniform loss of relative skill: tail performance depends on the model, variable, and evaluation setting rather than on whether the forecast is produced by AI or physical numerical modelling.