Microsoft's breakthrough in weather modeling

Extreme weather events are a growing challenge for communities and businesses. But now, we have a new tool to help us stay ahead of the storm. Microsoft researchers have developed #AuroraAI, a groundbreaking AI foundation model that forecasts not just weather, but a whole spectrum of environmental events – from hurricanes to air quality – with unprecedented accuracy.    Why is Aurora special? Here are a few highlights:  ✅ Smarter forecasts: Aurora outperforms traditional models on 91% of key forecasting targets, learning from over a million hours of weather data. It can predict a typhoon’s path days in advance more accurately – in one case, correctly forecasting a Typhoon’s landfall 4 days early when others missed it. ✅ Faster insights: It generates forecasts in seconds instead of hours, making it ~5000× faster than conventional systems. Speed saves lives and helps businesses make quick decisions. ✅ Beyond weather: Not just rain and storms – Aurora can predict air pollution and ocean waves too, beating the best models in those areas as well (it even improved forecasts of 5-day cyclone tracks beyond what any major center achieved).    Innovations like Aurora will help organizations and communities, as more precise and faster forecasts result in better decisions. Imagine farmers getting early warnings of a drought, or cities planning disaster response with AI guidance. Businesses can protect supply chains and build resilience against climate impacts. It’s a great example of technology empowering us to solve real-world challenges.    I’m especially proud that we’ve made Aurora source code and model weights publicly available – allowing researchers, startups, and institutions around the world to build on it. It’s currently powering more accurate forecasts on MSN Weather for everyone, and this is just the beginning. When I think about AI’s potential, one phrase comes to mind: “the defining technology of our time” – and innovations like Aurora prove it. 🚀    How do you see AI changing the way we forecast and plan?  Feel free to share your thoughts or learn more about Aurora AI (link in comments). Together, let’s harness these breakthroughs responsibly to create a more sustainable and secure future.    #AI #Innovation #ClimateTech #MicrosoftResearch #Sustainability 

🚨New open-source contribution to Aurora weather forecasting! Akshat Kaushal, a Masters student at Penn Engineering, has completed a comprehensive JAX implementation of Microsoft's Aurora foundation model for Earth system prediction. This work provides the research community with: • Complete training pipeline and optimization strategies • Detailed performance benchmarks comparing JAX vs PyTorch • Full code for model architecture, fine-tuning, and evaluation • Implementation of LoRA for parameter-efficient adaptation The JAX version enables enhanced customizability and leverages XLA compilation for potential performance gains, while maintaining numerical equivalence with the original PyTorch implementation. This contributes to democratizing access to advanced Earth system modeling tools and supports reproducible research in AI-driven Earth system science. 📄 Blog post: https://lnkd.in/dvP9U9wJ 💻 Code: https://lnkd.in/dpFr2mW6 #EarthSystemScience #MachineLearning #WeatherForecasting #OpenScience

Nature has just published Microsoft Research's Aurora, the first foundation #model of the #earth system. Aurora outperforms operational #forecasts in predicting #air quality, #ocean waves, tropical #cyclone tracks and high-resolution weather, all at orders of magnitude lower computational cost. Aurora first learns how to generate forecasts through training on #weather patterns from over one #million hours of data. These data are derived from satellites, radar and weather stations, simulations, and forecasts. The model can then be fine-tuned to perform a variety of specific tasks such as predicting wave height or air quality. When #Typhoon Doksuri hit the Philippines in July 2023, the damage was devastating. As reported in Nature, Aurora accurately predicts Typhoon Doksuri’s landfall in the Philippines using measurements from four days in advance of the event (image below). Official predictions at that time mistakenly placed the storm off the coast of Northern Taiwan. Results like this show how #AI is paving the way toward democratizing high-quality climate and weather prediction.  Learn more here: https://lnkd.in/gNiM5tsQ Try it here: https://lnkd.in/gn9DZsry

Microsoft’s 𝗔𝘂𝗿𝗼𝗿𝗮 𝘄𝗲𝗮𝘁𝗵𝗲𝗿 𝗳𝗼𝗿𝗲𝗰𝗮𝘀𝘁𝗶𝗻𝗴 𝗺𝗼𝗱𝗲𝗹 is driving actionable insights for the #energy sector and the communities we serve. Aurora can predict everything from hurricanes and typhoons to air quality and ocean waves with 𝗺𝗼𝗿𝗲 𝗽𝗿𝗲𝗰𝗶𝘀𝗶𝗼𝗻, 𝗳𝗮𝘀𝘁𝗲𝗿 𝘀𝗽𝗲𝗲𝗱, 𝗮𝗻𝗱 𝗮𝘁 𝗮 𝗳𝗿𝗮𝗰𝘁𝗶𝗼𝗻 𝗼𝗳 𝘁𝗵𝗲 𝗰𝗼𝗺𝗽𝘂𝘁𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗰𝗼𝘀𝘁 𝗰𝗼𝗺𝗽𝗮𝗿𝗲𝗱 𝘁𝗼 𝘁𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝗺𝗲𝘁𝗵𝗼𝗱𝘀 𝗮𝗻𝗱 𝗽𝗿𝗶𝗼𝗿 𝗔𝗜 𝗮𝗽𝗽𝗿𝗼𝗮𝗰𝗵𝗲𝘀. That’s a big deal when you’re trying to keep grids resilient and people safe. Energy runs on weather. Industry reports, like IEA, show 𝗯𝗲𝘁𝘁𝗲𝗿 𝘀𝘂𝗽𝗽𝗹𝘆-𝗱𝗲𝗺𝗮𝗻𝗱 𝗳𝗼𝗿𝗲𝗰𝗮𝘀𝘁𝗶𝗻𝗴 𝗰𝗮𝗻 𝗰𝘂𝘁 𝗿𝗲𝗻𝗲𝘄𝗮𝗯𝗹𝗲 𝗰𝘂𝗿𝘁𝗮𝗶𝗹𝗺𝗲𝗻𝘁 𝗯𝘆 𝘂𝗽 𝘁𝗼 𝟮𝟬%, unlocking cleaner and more dependable power for millions around the world. Huge congratulations to Juan M. Lavista Ferres, Christopher Bishop, and the Microsoft Research team for sharing this #AI breakthrough with the global scientific community and customers and partners everywhere. 🔗 Learn more about Aurora and stay tuned for how these models will power the next wave of #innovation for the energy sector. https://lnkd.in/gTj_2h8r #EnergyTransition #Sustainability

Microsoft Research just published a paper in Nature about Aurora, a foundation model trained on over a million hours of Earth system data. The idea is simple: one model, pre-trained on diverse geophysical datasets, then fine-tuned for specific tasks like air quality, wave forecasting, tropical cyclones or high-resolution weather. The results they report are strong, especially in benchmarks. Faster, lighter, and often more accurate than current operational models. But it's still early. These are "offline" experiments. Forecasting in real conditions is another story. You have to deal with multiple data sources, numerous forecast. This means choosing, based on knowledge of the quality of the different models and experience of each weather situation and microclimate... Still, it's a clear sign of where things are going. The gap between physics-based modeling and AI is closing, fast. Link to the paper in the first comment! #AI #Weather #Climate #Aurora #Forecasting

This work just published in Nature is impressive. The dominant large scale global dynamics are low dimensional and the cascade of those global features to local is also quite low dimensional. To get a fast and accurate weather model, the machine learning model might be latching onto the most important dynamics driving the weather systems. AI/ML is good for tackling large systems like this where underlying dynamics is inherently low dimensional. “With Aardvark, researchers have replaced the entire weather prediction pipeline with a single, simple machine learning model. The new model takes in observations from satellites, weather stations and other sensors and outputs both global and local forecasts. This fully AI-driven approach means that predictions are now achievable in minutes on a desktop computer. When using just 10% of the input data of existing systems, Aardvark already outperforms the United States national GFS forecasting system on many variables and it is also competitive with United States Weather Service forecasts that use input from dozens of weather models and analysis by expert human forecasters.”

Aurora, a 1.3 billion parameter foundation model from Microsoft Research AI for Science, offers high-resolution weather and air pollution forecasts by leveraging diverse atmospheric data. It efficiently adapts to new tasks, outperforms traditional models in extreme weather prediction, and delivers rapid results, all while being computationally efficient. For municipalities, it provides rapid and accurate predictions, aiding in better decision-making for urban planning, emergency response, and environmental management, while efficiently handling extreme weather events. Paper: https://lnkd.in/dMa6Tkya Github: https://lnkd.in/dwJYA3WY Infographics: Napkin AI #AI #ClimateScience #WeatherForecasting #MunicipalPlanning #UrbanManagement #EmergencyResponse #EnvironmentalManagement #MachineLearning #SmartCities #PublicSafety

How can Microsoft's Aurora AI change weather predictions? The topic matters because accurate weather forecasting affects everyone, from farmers to commuters. Microsoft's Aurora AI claims to predict atmospheric events with remarkable precision, shaking up traditional meteorological methods. 🌤️ Microsoft has unveiled its Aurora AI model, designed to predict air quality and severe weather events with impressive accuracy. This innovation is not just tech fluff; it’s aimed at revolutionizing how we prepare for storms and other weather-related phenomena. With its remarkable ability to forecast typhoons and hurricanes, Aurora could be a game changer in how we respond to natural disasters. 🌪️ Imagine weather forecasting as a high-stakes game of chess, where every move counts. Microsoft has trained Aurora on over a million hours of diverse data, making it swift and efficient in generating predictions that beat traditional methods. For instance, it successfully predicted Typhoon Doksuri's landfall four days early, showcasing its potential to save lives and resources. 💬 Interested in how AI is reshaping our understanding of weather? What do you think about the role of technology in disaster preparedness? Share your thoughts below! #Microsoft #AI #WeatherForecasting #AuroraAI #Technology #Innovation #DataScience #Meteorology #ClimateChange

I have been working in the AI world for quite a while, and nothing makes me more excited than seeing AI’s potential to transform our future for the collective good. As a very early leader and proponent in the AI for Good program at Microsoft, I have had the privilege of launching impactful initiatives and campaigns that have driven our mission forward. I am also honored to work closely with Microsoft Research. Their dedication and innovation in the field of AI has been a never-ending source of inspiration. When we talk about AI’s potential to transform our future for good, the kind of work below is what I’m talking about. Congratulations to the Microsoft Research team on this incredible achievement!   A new paper published in Nature demonstrates how Aurora, an AI foundation model developed by Microsoft Research, can predict a wide range of atmospheric events with greater precision and accuracy than traditional numerical methods and existing AI models. Built on flexible architecture using more than one million hours of diverse atmospheric data, Aurora doesn’t just forecast weather patterns – it can predict air quality, hurricanes and typhoons, the height of ocean waves, even sandstorms.   In the hands of climate researchers, meteorologists, and other scientists, this technology could be leveraged to enhance crop logistics, protect energy grids, support public health efforts, and help industries and governments better prepare for extreme climate events.     Learn more: https://lnkd.in/gDp8YEjS   #AI #MicrosoftResearch #Forecasting

The weather and AI ☂🤖 Talking about the weather is not just a recurring way to start a conversation or break an awkward silence. Predicting it accurately and timely is increasingly key for determining the availability of renewable energy, planning routes, or effectively managing investments in agriculture or infrastructure. And, even more importantly, for saving lives. Yet, anticipating atmospheric evolution is an extremely complex process involving hundreds of erratic factors. Alterations in any of them can ruin a forecast. Because of this, the most reliable predictions do not last more than three days. But in this case (as in many others), AI can help. #NatureMagazine has just published an article on Aurora, an AI model developed by #MicrosoftResearch in collaboration with several universities that outperforms previous prediction systems at a fraction of the computational cost. At #Equinor, we are also developing AI tools to improve weather forecasts and reduce emissions at our offshore facilities: Hywind Tampen is a great example. 🌊 Undoubtedly, AI raises several questions and dilemmas. But it is also true that its responsible use allows safer, more efficient, affordable and accessible tools. As the authors in Nature state, "Aurora represents a notable step towards democratizing accurate and efficient Earth system predictions." https://lnkd.in/eHgDfFNG 👈🏻