.png)
On Wednesday, Google unveiled GenCast, a weather-predicting artificial intelligence (AI) model. Google DeepMind, the internet giant's AI research branch located in Mountain View, created the AI model. The company's researchers recently released a study on the system, showcasing its ability to make medium-range weather forecasts. The business says that the system outperformed current state-of-the-art forecasting methods in terms of resolution and accuracy. Notably, GenCast can forecast the weather for the following 15 days with a resolution of 0.25 degrees Celsius.
Google Gencast Features
In a blog post, Google DeepMind described the new high resolution AI ensemble model. GenCast, which can predict day-to-day weather and severe occurrences, claimed to outperform the European Centre for Medium-Range Weather Forecasts' (ECMWF) Ensemble (ENS) algorithm. The model's performance has recently been reported in the journal Nature.
Notably, GenCast takes a probabilistic approach to weather prediction rather than the standard deterministic one. Weather prediction models that use the deterministic method generate a single, particular forecast for a given set of beginning circumstances and are based on accurate atmospheric physics and chemical equations.
Probabilistic models, on the other hand, create several probable outcomes by simulating a variety of beginning conditions and model parameters. This is also known as ensemble forecasting.
Google DeepMind described GenCast as a diffusion model that adapts to the Earth's spherical topology and develops complicated probability distributions for future weather events. To train the AI model, researchers used four decades of historical weather data from the ECMWF's ERA5 collection. With this, the model was taught global weather patterns with a resolution of 0.25 degrees Celsius.
In the released research article, Google assessed GenCast's effectiveness by training it on historical data up to 2018 and then requesting forecasts for 2019. A total of 1320 combinations across different variables at varying lead times were examined, and the researchers discovered that GenCast outperformed ENS on 97.2% of these goals, and 99.8% for lead periods larger than 36 hours.
Notably, Google DeepMind has stated that it would make the code, weights, and forecasts for the GenCast AI model available to the weather forecasting community.
#google #AI #weatherforecast #gencast #aipoweredweatherforecasting #newstainmentora
