Google unveiled a potentially groundbreaking development in weather forecasting. The company’s latest innovation, known as its “Scalable Ensemble Diffusion Sample” (SEEDS), is using cutting-edge artificial intelligence to revolutionize weather forecasting.
Similar to language models like ChatGPT, SEEDS taps into the power of AI to churn out forecasts in record time. SEEDS is different due to its ability to dig deeper into weather extremes. The technology can detect dangerous events that may slip through the cracks of traditional weather models.
Scientists say that with climate change, extreme weather events are becoming more frequent and making accurate and timely forecasts is more vital than ever before.
SEEDS is also cheaper than traditional weather models. While traditional models use a wide range of forecasts to predict 10-50 weather scenarios, SEEDS needs just one or two forecasts to predict 31 different weather scenarios.
Another difference between SEEDS and traditional weather models is that traditional models use a set of initial conditions and random variables to generate forecasts, like humidity, temperature, wind speed and air pressure. However, traditional methods often struggle to accurately predict extreme weather as atmospheric dynamics change.
In contrast, SEEDS approach focuses on the relationships between different weather variables such as energy and sea level pressure. The approach allows for more detailed forecast models, providing a more thorough picture of future weather conditions.
For example, SEEDS played a crucial role in predicting Europe’s 2022 heatwave. Just seven days before the heatwave hit, traditional forecasting methods showed no indication of the extreme temperatures. However, SEEDS predicted the event, providing advanced warnings and allowing people to plan, possibly saving lives.