Microsoft introduces its AI forecast, Aurora, capable of detecting storms weeks ahead

After Storm Ciarán’s surprising strength caused destruction in northwestern Europe, it became clear that existing weather prediction models have limits.

This event was a rare meteorological occurrence for England, and it highlighted the pressing requirement to improve forecasting methods so as to effectively prepare for and lessen the impacts of severe weather occurrences like this one.

Now comes Aurora, Microsoft’s pioneering AI-based weather forecast model. According to the Redmond-based tech giant, it has been carefully created to change how we predict the weather using large quantities of atmospheric data, helping us foresee and anticipate the intense increases in events similar to Storm Ciarán.

This flexible 3D Swin Transformer model with 1.3 billion parameters has been fine-tuned to provide high-resolution forecasts of weather and atmospheric processes.

The model consists of a flexible 3D Swin Transformer with Perceiver-based encoders and decoders, enabling it to process and predict a range of atmospheric variables across space and pressure levels. By pretraining on a vast corpus of diverse data and fine-tuning on specific tasks, Aurora learns to capture intricate patterns and structures in the atmosphere, allowing it to excel even with limited training data when it is being fine-tuned for a specific task. 

Microsoft

The special thing about Aurora is its capacity to learn from more than a million hours of different weather and climate simulations. This makes it very good at prediction work in places where data is less present or during severe weather situations.

This ability is very important, especially as climate change makes the weather increasingly unpredictable.

The model’s spatial resolution of 0.1°, or about 11 km at the equator, is very high. It can catch intricate information in the atmosphere. This means Aurora can predict more precisely than ever before and at a computational speed-up that might be around 5,000 times faster than normal forecasting systems. Think about how this could make weather predictions timely, efficient, and accurate like never before!

However, the AI-powered Aurora’s strength is not just in weather forecasting. It also shows its ability to predict atmospheric chemistry and air pollution levels, doing better than operational models for many targets.

It can be used in many ways because its complex structure can handle different inputs and make predictions at various resolutions and fidelity levels. Aurora’s capability to successfully detect variations in pollutants like nitrogen dioxide at extremes and background levels highlights its potential for enhancing air quality monitoring and prediction.

The special ingredient that makes Aurora so great is the pretraining it goes through with various datasets. This ensures Aurora performs well in many different tasks and resolutions, surpassing the best numerical weather prediction models and deep-learning approaches.

The model’s capacity to accurately predict extreme values reveals its potential for revolutionizing Earth system modeling.

The making of Aurora shows us that AI is very important to improve the general knowledge and forecasting of Earth’s intricate systems. Even though Microsoft already has some of the best weather forecasting systems in the world, the AI-powered Aurora might make it easier for everyone to get important weather and climate details worldwide, especially in places with less data available.

This can influence many areas, such as agriculture, transportation, energy, and disaster readiness. It may assist communities worldwide in adjusting to the difficulties brought by climate change.

In the time ahead, Aurora’s plan gives us thrilling chances to create detailed models that include the total Earth system. As computer resources and data keep improving, basic models such as Aurora might become the main structure of weather forecasts and climate prediction systems.

It means people who make decisions and the general public could get precise information at the right moments to make choices about how to respond actively to problems on our changing planet.

Microsoft has tested Aurora. The Redmond-based tech giant published a paper earlier this year describing a patented technology powered by AI that can predict storms. We thought it was Copilot at the time; it turns out it was the new Aurora AI weather forecasting model.