If you’ll be attending NeurIPS—or are interested in AI for weather and climate—feel free to reach out. I’d be very happy to connect!

@aifraunhoferhhi.bsky.social
/end

On the Predictive Skill of Artificial Intelligence-based Weather Models for Extreme Events using Uncertainty Quantification Accurate prediction of extreme weather events remains a major challenge for artificial intelligence based weather prediction systems. While deterministic models such as FuXi, GraphCast, and SFNO have ...

We’ve also released a pre-print, which not only includes this case study but also an extended analysis of the August 2022 China heatwave, offering a broader evaluation. 📄 Read the full pre-print here: arxiv.org/abs/2511.171...

3/

On the Predictive Skill of Artificial Intelligence-based Weather Models for Extreme Events using Uncertainty Quantification Accurate prediction of extreme weather events remains a major challenge for artificial intelligence based weather prediction systems. While deterministic models such as FuXi, GraphCast, and SFNO have ...

Using the 2022 Pakistan floods as a case study, we evaluate state-of-the-art deterministic AI models and examine how different perturbation methods can extend them toward trustworthy probabilistic forecasts for extreme precipitation, which is essential for future early warning systems.

2/

I’m pleased to share that next week I will be presenting at the @climatechangeai.bsky.social Workshop at NeurIPS 2025 in San Diego.

We investigate a central question: Can current AI-based global weather models reliably capture extreme events?

1/