As a global centre of excellence in weather and climate science, the Met Office has always been driven by innovation.
Since our founding in 1854, we have pushed the boundaries of meteorology, bringing together scientific expertise, cutting-edge technology and a commitment to providing trusted weather and climate services. Today, artificial intelligence (AI) and machine learning (ML) represent the next major leap forward in forecasting capability.
These technologies offer transformative potential across every stage of our work, from understanding the atmosphere to delivering actionable insights for government, businesses and the public.
Through our National Capability AI Programme (NCAIP) and wider research and development efforts, we aim to deliver significant benefits across our work with government, businesses and the public. By combining physics‑based modelling, expert meteorological knowledge and the power of data science, we aim to maintain the UK’s leadership in weather and climate modelling and deliver more accurate, efficient and trusted intelligence.
How AI supports our mission
Weather and climate science rely on enormous volumes of data: satellite imagery, surface observations, radar scans, ocean measurements, climate analyses and more. They also depend on sophisticated models that simulate the behaviour of the Earth system. Working with this data and these models requires significant computational resources and complex workflows stretching from observation to simulation, interpretation and service delivery.
AI and ML offer new ways to navigate this complexity. These technologies can help analyse data more efficiently, improve model performance and enhance the products and services we provide. Crucially, AI does not replace traditional science; instead, it can complement our physics‑based models and expert insight by adding powerful new tools for discovery, simulation and decision‑making.
Our data science framework sets out how we can integrate AI responsibly and effectively, focusing on trustworthiness, interpretability and scientific rigour. Alongside technological innovation, we continue to invest in skills, culture and partnerships to ensure our people and collaborators can make the most of these new capabilities.
READ MORE: Can AI developments really be green?
AI and Met Office capabilities
We are applying AI and ML across several key areas of our work:
- Discovery and attribution
AI can support data mining and pattern discovery in vast datasets, helping us detect, quantify and better understand the causes of weather and climate phenomena. This includes uncovering relationships that may not be immediately apparent and testing hypotheses more efficiently.
- Fusing simulation with data science
AI can be used to replace or enhance specific components of numerical weather prediction (NWP) or climate models. This may reduce model run times or improve accuracy, helping to advance forecasting capability.
- Uncertainty and trust
Ensuring that AI‑driven outputs are reliable, explainable and scientifically sound is a priority. We are developing methods to interpret the predictions made by machine‑learning systems so meteorologists and decision‑makers can understand, evaluate and trust them.
- Data to decisions
AI could help refine outputs from NWP and climate models, tailor insights to user needs and provide information through improved interfaces. This could support better decision‑making across sectors such as aviation, energy, emergency response and environmental management.
Together, these capabilities illustrate the breadth of opportunity AI brings to weather and climate science, supporting more agile innovation, greater efficiency and deeper insight.
AI in weather science
Advances in data science now allow computers to analyse and learn from unprecedented volumes of information. This presents a significant opportunity to enhance the speed, accuracy and usefulness of weather forecasting.
As the UK’s national meteorological service, we are accelerating the safe adoption of AI and ML across our operational and research activities. Through pilot projects, collaborations and capability development, we are exploring how these technologies can augment the forecasting process from end to end in the coming years.
Why AI forecasts are needed
Extreme weather events are becoming more frequent and intense as a result of climate change. Improving our ability to understand and predict these events is vital to helping society stay safe and resilient. AI‑driven modelling offers a timely opportunity to advance forecasting capability further, complementing decades of progress in physics‑based NWP.
How AI could enhance forecasts
AI has the potential to strengthen the forecasting process at multiple stages:
- Observations: Quality control, error detection and filling in gaps in observational data
- Simulation: Enhancing data assimilation or replacing parts of physical models
- Analysis: Improving post‑processing and refinement of forecasts
- Services: Developing risk‑based products and user‑specific insights
These techniques can work alongside existing models or offer entirely new approaches to predicting the weather.
Physics‑based and AI‑based models
Existing physics-based weather prediction relies on solving equations that describe atmospheric physics. Our operational Momentum Model, for example, uses supercomputers to simulate how the atmosphere evolves over time.
AI‑based models take a different approach. Models such as FastNet, in development between the Met office and the Alan Turing Institute, are demonstrating the potential in this area.
Once trained, AI models can produce forecasts in a fraction of the time required for a full physics‑based simulation. This opens potential for higher resolution predictions, reduced computational costs and rapid updates.
However, AI models do not yet fully capture the underlying physics, which means they may struggle with rare or extreme events. For this reason, future forecasting systems are likely to blend both approaches, with meteorologists interpreting and integrating outputs to produce a trusted forecast.
Supercomputers will remain essential, not only for running physics‑based models but also for generating the datasets needed to train AI.
READ MORE: How will AI weather forecasts make maximum impacts for users?
The continuing role of meteorologists
Despite the power of AI, expert human interpretation will remain indispensable. Meteorologists play a key role in validating model output, assessing uncertainty and providing clear, trusted guidance. Just as computers transformed forecasting in the 1960s, AI offers a new layer of capability - but professional judgement continues to be vital.
We are supporting our workforce to use AI effectively, ensuring our organisation remains resilient, efficient and ready for the challenges ahead.
AI in climate science
AI and ML could shape climate science in similar ways to weather forecasting. Climate models represent long‑term changes in the Earth system and traditionally improve through better science and increased computational power. AI offers new routes to enhance small‑scale processes, improve localised impact projections and support analysis of extreme events.
Through the AI4Climate programme, funded by the Department for Science, Innovation and Technology, we aim to explore cutting‑edge AI methods to strengthen climate modelling and deliver improved climate information more efficiently. This work supports our broader aim of combining simulation, data science and AI to deliver world‑leading climate intelligence.
Keep up to date with weather warnings, and you can find the latest forecast on our website, on YouTube, by following us on X and Facebook, as well as on our mobile app which is available for iPhone from the App store and for Android from the Google Play store.
