Integrating physical laws into machine learning improves weather and climate forecasting

Nature · Published 2 October 2026 Swipe more like this →

In short: Researchers who surveyed various machine learning applications in meteorology found that standard models often fail because they ignore the fundamental laws of physics. By systematically incorporating domain knowledge and physical constraints into these algorithms, scientists have successfully improved how we emulate, downscale, and forecast complex atmospheric processes.

While off-the-shelf machine learning is powerful at recognizing patterns, it can produce results that are physically impossible or fail in new scenarios. This study reviewed 10 specific case studies to show how "physics-informed" models achieve better data efficiency and physical consistency compared to traditional methods.

The review identifies several remaining challenges, including scientific, diagnostic, and computational hurdles. Because this is a survey of existing literature rather than a new experimental trial, it provides a broad overview of current progress rather than new raw data.

Why it matters: Combining the pattern-recognition power of AI with the reliable rules of physics creates more accurate climate models.

General information, not professional advice. Spotted an error? Flag the card or see corrections and complaints.

Intellect Plus is a visual educational discovery platform for exploring fascinating facts and ideas. Brain Food, Not Brain Rot. What is Intellect Plus?

Related

• Can a machine learning model outperform massive, expensive weather forecasting supercomputers?
• How supercomputers turn current weather observations into tomorrow's forecast using complex mathematical models.
• Generative AI improves the accuracy of short-term rainfall forecasts
• Why your weather forecast relies on physics happening at scales too small to see
• Efficient neural network architecture achieves accurate short-term precipitation forecasting
• Key barriers and training needs for agricultural extension agents in northeastern Ghana
• Integrated crop and livestock systems may boost food security and sustainability
• Key factors influence how Ugandan farmers adapt to climate change
• Industrial F-gases are rapidly growing super pollutants with massive warming potential
• Land and sea use change are the primary drivers of global biodiversity loss
• Indian farmers adapt to climate change through incremental and systemic shifts
• Large-scale climate simulations help distinguish natural variability from human-driven change

More in Nature · Latest · A–Z · All topics