An artificial intelligence model, developed by Microsoft, led by the Greek professor at the University of Pennsylvania, may successfully predict different aspects of Earth’s behavior, from weather patterns to air pollution and waves.

The model is called Aurora and is presented in a publishing of Nature magazine.

How it works

The prognosis of Earth systems serves as an integral tool for providing timely warnings for extreme events. The forecasts come from complex models, which are based on decades and often require supercomputers and specialized groups for their maintenance, thus becoming not accessible to many communities worldwide.

Aurora is an artificial intelligence model trained in over a million hours of geophysical data. The purpose of its creation was twofold: to create a prediction tool that is at the same time more expensive and computational much more efficient. Adopting a fundamentally different approach from traditional weather forecast models, Aurora learns patterns directly from data by identifying complex relationships in historical data of the earth and uses them to make predictions.

As noted in the publication, Aurora exceeds existing models for air quality, ocean waves, tropical cyclone routes and high -resolution weather forecastat a lower computing cost than current prediction methods. According to the data listed, for predicting air quality, Aurora reached or exceeded Copernicus Atmosphere Monitoring Service in 74% of the targets, and was about 50,000 times faster. In addition, for high -resolution weather conditions, the model exceeded the performance of the top IFS HRS weather model in 92% of the 0.1 ° resolution targets, performing better performances in extreme events.

“Aurora represents an important innovation in environmental systems prediction, as it is the first artificial intelligence model that works as single fundamental model Able to adapt to different applications, from high -resolution weather forecast and air quality predictions to monitoring of tropical cyclones and ocean waves. This approach achieves high accuracy with thousands of times lower computing costs, making the advanced environmental forecasts in wider communities worldwide, ”Mr Perdikaris explains to the Athens-Macedonian News Agency.

The main innovation of the model is the fact that can be trained in a huge volume of diverse geophysical data and then optimized for specific prediction work as a powerful brain that can be specialized to perform different forecasting.

As Mr Perdikaris observes, “the Aurora project, during my term in Microsoft Research, was part of my broader research vision to create fundamental models for scientific applications that can be generalized in various areas and accelerate discoveries.” He adds that at the University of Pennsylvania “my team expands this vision beyond the Earth’s sciences in a variety of science and engineering applications, creating artificial intelligence systems that not only can predict but also help us understand complex natural phenomena.” Similar approaches to models also apply to other scientific areas, from materials to biomedical applications.