cloudLookbook
AI Research

GenCast

2024arrow_outward

GenCast: Revolutionizing Weather Forecasting

GenCast is an advanced AI-based weather model from Google DeepMind that predicts weather and the risks of extreme conditions with state-of-the-art accuracy up to 15 days ahead. It represents a paradigm shift from traditional physics-based models to data-driven generative AI.

Extreme Event Prediction

Provides enhanced forecasts for extreme weather events like heatwaves, high winds, and heavy precipitation with unmatched accuracy compared to existing models.

State-of-the-art Accuracy

Outperforms top traditional ensemble models such as ENS globally across a wide range of weather variables and lead times.

How GenCast Outperforms the Rest

1

Generative AI Core

Uses a machine learning approach trained on decades of historical weather data (like ERA5) rather than relying solely on solving complex fluid dynamics equations.

2

Probabilistic Forecasting

Generates ensembles of possible future weather scenarios, providing crucial probabilities for rare but impactful extreme weather events.

3

Computational Efficiency

Produces 15-day ensemble forecasts on a single TPU in just a few minutes, orders of magnitude faster and more energy-efficient than supercomputer-based traditional models.