AI‑Boosted Subseasonal Forecasts Double Predictive Skill, Offering New Climate‑Planning Tools

Researchers developed a probabilistic bias correction machine-learning method that doubles the predictive skill of subseasonal weather forecasts. This approach corrects systematic errors in both AI and physics-based models to improve temperature, precipitation, and pressure accuracy.

The two-to-six-week subseasonal window has historically been difficult to predict, often failing to outperform simple climatological averages. This new tool addresses these gaps to assist sectors like agriculture and energy in planning for extreme weather events.

The new method marks the first time subseasonal forecasts consistently outperform climatological averages. Unlike resource-heavy models, this approach runs on modest hardware, increasing accessibility for global weather services.