Source: TH
Context: Google DeepMind recently unveiled GenCast, a groundbreaking AI-based weather forecasting model.
About GenCast:
- What is GenCast?
-
- GenCast is a diffusion-type AI model designed for probabilistic weather forecasting, predicting weather conditions using machine learning techniques.
- Parent Company: Developed by Google DeepMind.
- How it works:
-
- Uses ensemble forecasting: Generates multiple predictions by combining historical data with noisy inputs and refining them iteratively through neural networks.
- Trained on 40 years of reanalysis data (1979-2019).
- Produces forecasts for up to 15 days with a spatial resolution of 0.25° x 0.25° and temporal resolution of 12 hours.
- Existing forecast models:
-
- Numerical Weather Prediction (NWP): Relies on solving physical equations but requires high computational power and provides deterministic forecasts.
- Huawei’s Pangu-Weather: Predicts weekly weather faster than NWP models.
- Superiority of GenCast:
-
- Probabilistic Forecasts: Better at predicting extreme weather and providing longer lead times for disaster preparation.
- Efficiency: Faster and more resource-efficient than NWP models.
- Extreme Event Prediction: Superior in tracking tropical cyclones and wind power production.
Insta links:
