Microsoft has upgraded its powerful weather AI model, Aurora. It can now predict air quality as well. The announcement was made in a blog post earlier this week. Aurora was created by Microsoft Research. It can forecast a wide range of weather events like hurricanes and typhoons. The model is said to be faster and more accurate than traditional methods.
Microsoft has also published a research paper on Aurora in the science journal Nature. The company has made Aurora’s source code and model weights public. A version of the model, designed for hourly weather updates, has been added to the MSN Weather app. This version can forecast clouds as well.
Microsoft claims Aurora is one of the best AI weather models in the world. The company said it is trained as a foundational model. This means it learns from large datasets first, and then can be fine-tuned for specific tasks. One such task is predicting air pollution. Microsoft said that because Aurora has already learned from vast amounts of weather data, it only needs a small amount of air quality data to specialise further.
Also read: India’s fully indigenous drone fleet
The model was trained using more than a million hours of satellite, radar, and weather station data. It also used past weather simulations and forecasts. Aurora uses an encoder architecture. This allows it to process huge amounts of data and convert it into a format the AI can understand. It doesn’t rely on pre-set rules. Instead, the model learns from data patterns on its own.
“We’re not putting in strict rules about how we think variables should interact with each other,” said Megan Stanley from Microsoft Research. “This is the power of deep learning in these kinds of simulation problems.” Aurora has already shown strong results. It correctly predicted Typhoon Doksuri’s landfall in the Philippines four days in advance. It also forecast a sandstorm in Iraq two years ago. In 2022 and 2023, it outperformed the US National Hurricane Center in predicting tropical cyclone paths.
Unlike traditional weather systems that rely on supercomputers, Aurora runs on graphics processing units (GPUs). This allows it to deliver forecasts within seconds. Microsoft noted that while training Aurora was expensive, it is cheaper to operate than older forecasting systems. AI models for weather are not new. Google DeepMind has developed several, including WeatherNext. But Microsoft’s Aurora is now leading the race for faster and smarter weather predictions.
Follow us
