weather forecasts

Microsoft this week unveiled an AI foundation model that’s designed to create highly accurate predictions not only for extreme weather events but also for other atmospheric dynamics, such as temperature, wind speed, air pollution levels, and concentrations of greenhouse gases.

Data generated by the Aurora model could help governments, communities, and businesses better prepare for potentially danger weather events in a rapidly evolving climate environment, according to researchers at Microsoft.

They pointed to Storm Ciarán, a low-pressure system that slammed Europe in the first days of November 2023, surprising experts with rainfall and wind speeds that exceeded their predictions and killed 21 people across France, Italy, and other countries.

“The storm’s intensity caught many off guard, exposing the limitations of current weather-prediction models and highlighting the need for more accurate forecasting in the face of climate change,” they wrote in a blog post. “Aurora presents a new approach to weather forecasting that could transform our ability to predict and mitigate the impacts of extreme events – including being able to anticipate the dramatic escalation of an event like Storm Ciarán.”

Aurora is a 1.3 billion parameter foundation model built for high-resolution forecasting of weather and atmospheric processes and trained on more than a million hours of varied weather and climate simulations to give it what the researchers called a “comprehensive understanding of atmospheric dynamic. This allows the model to excel at a wide range of prediction tasks, even in data-sparse regions for extreme weather scenarios.”

It also can do so quickly. Microsoft said the foundation model can bring about 5,000 times the computational speed than a traditional numeric weather-prediction system at a much lower cost. In less than a minute, Aurora can produce a five-day global air pollution prediction or a 10-day high-resolution weather forecast.

“The ability of foundation models to excel at downstream tasks with scarce data could democratize access to accurate weather and climate information in data-sparse regions, such as the developing world and polar regions,” the researchers wrote. “This could have far-reaching impacts on sectors like agriculture, transportation, energy harvesting, and disaster preparedness, enabling communities to better adapt to the challenges posed by climate change.”

AI and the Weather

Weather forecasting has become a hot spot for AI and a global weather forecasting services market that is expected to grow from $2.14 billion in 2021 to $3.78 billion by 2030, according to analysts with market research firm Straits Research.

“AI, with its remarkable ability to sift through immense datasets to uncover complicated patterns, heralds a new era in meteorology,” Rosella (Qian-Ze) Zhu, a PhD student at Harvard’s Kenneth C. Griffin Graduate School of Arts and Science, wrote in a blog post in March. “From providing farmers with precise agricultural forecasts to predicting the path of deadly cyclones, AI and [machine learning] are transforming how we interact with and understand the weather.”

Tech giants are rolling out their own AI models for weather forecasting. They include Google and its GraphCast, NetMet-3, and SEEDS AI models, Nvidia with FourCastNet, and Huawei with Pangu-Weather. IBM and NASA last year used Big Blue’s watsonx AI platform to develop a climate foundation model.

Investors also are eyeing the industry. IBM last year sold The Weather Company to private equity firm Franciso Partners and in February startup Jua raised $16 million in investments to develop an AI large physics model for predicting weather. Windborne Systems, a startup developing AI-based weather balloons, this week announced a $15 million investment round led by Khosla Ventures.

“AI is fundamentally changing weather prediction, which hasn’t been meaningfully disrupted since the 1990s but is essential to better understand and address the impacts of climate change,” Sven Strohband, a partner at Khosla Ventures, said in a statement.

Varied Pretraining Data a Key

Microsoft’s researchers see Aurora as both a highly accurate and fast atmospheric forecasting system and an agent of change, saying they “hope Aurora will serve as a blueprint for future research and development. The study highlights the importance of diverse pretraining data, model scaling, and flexible architectures in building powerful foundation models for the Earth sciences.”

The model operates at a high spatial resolution – about 11 km at the equator – enabling it to collect intricate details of atmospheric processes. It also leverages its diverse sets of pretraining data and encoder and decoder architecture to develop its highly versatile capabilities, from looking at temperatures to where greenhouse gases are collecting.

The researchers pointed to the model’s capabilities forecasting air pollution levels using data from the Copernicus Atmosphere Monitoring Service (CAMS), which they called a “notoriously difficult task due to the complex interplay of atmospheric chemistry, weather patterns, and human activities, as well as the highly heterogeneous nature of CAMS data.”

Aurora can process and learn from the data and capture the unique characteristics of air pollutants and their relationships with meteorological variables, they wrote. The model can product an accurate five-day global air pollution forecast and outperform current atmospheric chemistry simulations on 74% for the targets, even in scenarios where there isn’t a lot of data or the environment is highly complex.