A partnership between the Turing and the Met Office aims to make weather forecasting more accurate and efficient write the Met Office's Professor Kirstine Dale and Dr Scott Hosking, senior research fellow at the Alan Turing Institute.
Our lives revolve around the weather. Checking the forecast helps us decide how to commute to work, when to hang the washing out and what to do at the weekend. Perhaps we don't think too deeply about what is involved in producing accurate forecasts. But our changing climate is causing more frequent and intense weather extremes and, as artificial intelligence (AI) algorithms advance at pace, scientists are starting to reimagine the weather forecast for a more volatile future.
The Alan Turing Institute is collaborating with the UK Met Office – a leading force in weather forecasting for over 160 years – to explore how AI and data science can enhance and even transform weather forecasting. This effort forms part of the Turing's Environment and Sustainability Grand Challenge (announced earlier this year), and it has the potential to benefit communities and businesses worldwide.
Why do we need to take weather forecasting to the next level?
Weather forecasting doesn't just dictate our daily lives. Accurate weather predictions, for example, help governments, emergency responders, the public and businesses prepare for extreme events such as storms, floods and heatwaves. Forecasts also inform decision-making in weather-sensitive sectors, such as choosing harvesting times to improve crop yields or optimising national energy grids to make best use of renewable energy capacity.
A 2023 study estimated the global cost of extreme weather events in the past 20 years at £13m an hour, whilst UNICEF recently reported that extreme weather displaced 43m children over the past six years. This year has been another record-breaking year of extreme weather: flooding in South Korea, China, Greece and Libya, wildfires in Canada and around the Mediterranean, heatwaves across Europe, and unusually cold temperatures in Afghanistan. More recently, Storm Babet brought intense rainfall and winds to Europe, causing significant flooding across parts of the UK. Clearly, there is a need to make weather forecasting as accurate, computationally efficient and cost-effective as possible.
How does existing weather forecasting work?
Traditional weather forecasting relies mostly on complex numerical weather prediction (NWP) models – often consisting of millions of lines of computer code – which simulate the Earth's atmosphere using mathematical equations. These models require vast amounts of data, including atmospheric measurements, satellite observations and historical weather data, as well as powerful computational resources.
While NWP models have significantly improved over the years, they can struggle with forecasting local events such as showers, making predictions less accurate over certain regions and for specific events. Moreover, as more satellite sensors become available, it can be time-consuming to integrate new datasets of various modalities, resolutions and scales.
How does AI weather forecasting differ?
Rather than building modelling systems bottom-up using known physical principles, AI uses historical data to learn how weather patterns evolve. Machine learning algorithms are trained on vast quantities of existing data – including data from various satellites, surface sensors and the outputs of physics-based NWP models – to automatically recognise complex patterns in weather phenomena. These AI weather models can already compete with the current generation of NWP models and, for some applications, exceed their performance at a fraction of the computational cost.
Big tech companies including Google, Microsoft, Nvidia and Huawei have made significant advances in applying AI to weather prediction over the past two years. This has taken the international weather forecasting community by storm (pun intended), clearly demonstrating the potential of AI-powered weather models to complement traditional physics-based NWP models, especially for short-term forecasting. As AI algorithms and methods are developing so rapidly, we can expect to see even more advancements in the coming years, both within and outside the big tech labs. The low cost of producing AI forecasts means that they have the potential to democratise forecasting for the wider research community, benefitting people all over the world.
Our partnership presents an exciting opportunity to combine the Turing's AI and data science expertise with the Met Office's world-leading forecasting capabilities. Together, we aim to address some of the technical and logistical challenges that still face AI weather forecasting, such as:
- Data quality: High-quality, comprehensive data is essential for training AI models effectively. Ensuring data accuracy and availability is a priority.
- Model complexity: Developing AI models that can handle the intricacies of atmospheric science while remaining computationally efficient is a significant challenge.
- Integration: Integrating AI-based forecasting seamlessly into existing meteorological systems and workflows is crucial for practical implementation.
- Validation and verification: Rigorous testing and validation are needed to ensure the reliability and accuracy of AI weather forecasts.
Possible outcomes from our partnership include more accurate and timely prediction of extreme weather events (helping to save lives and protect critical national infrastructure); a greater breadth of user-focused weather forecasting services to aid decision-making across the public sector and private industry; and localised forecasts that can provide weather information at ever finer resolution.
Ultimately, our goal is to benefit individuals and industries, contributing to the global effort to build resilience to weather extremes. As AI continues to evolve and integrate into our lives, the outlook for the future of weather forecasting looks brighter than ever.
For more information on the first phase of our partnership, read our news release.
No comments:
Post a Comment