AI Insight
This study demonstrates that probabilistic AI weather prediction models outperform traditional physics-based numerical weather prediction models when forecasting extreme heat events, including record-breaking temperatures. The researchers evaluated global land forecasts and found that AI models better predicted both the occurrence and intensity of heat extremes compared to the ECMWF IFS ensemble across all regions and lead times. The authors argue that previous reports showing AI model limitations were due to inappropriate verification methods, specifically the comparison of deterministic forecasts rather than probabilistic assessments.
Why it matters
Accurate prediction of extreme heat events is critical for public health preparedness, agricultural planning, and infrastructure management. This research suggests that AI weather models, when properly evaluated using probabilistic methods, may provide superior early warning capabilities for dangerous heat waves that break historical records.
Understand the Science
⚠️ Preprint – Noch nicht peer-reviewed
Dieser Artikel wurde noch nicht von unabhängigen Experten begutachtet. Die Ergebnisse sind vorläufig und sollten mit Vorsicht interpretiert werden.
Abstract: AI weather prediction (AIWP) models have emerged as a key complement to physics-based numerical weather prediction (NWP) models, and now often outperform them on aggregate skill. Yet, several studies report that physics-based models still forecast extreme events better, in particular record-breaking extremes, based on the evaluation of deterministic forecasts. Here we evaluate heat extreme forecasts from probabilistic AIWP models over global land and show that they clearly outperform the ECMWF IFS ensemble, both raw and post-processed, including for record-breaking events, in all regions and at all lead times considered. AIWP models better forecast the occurrence and intensity of extremes, are better calibrated in the tail and miss fewer records. Station observations confirm these results. We argue that many reported AIWP limitations stem from unsuitable verification choices, specifically comparing deterministic forecasts that target different quantities. Instead, a fair evaluation of extremes necessitates probabilistic verification.