The Met Office stands at the forefront of weather and climate science, continually evolving its capabilities to deliver ever more accurate and useful forecasts.
As we approach a new era in supercomputing, the Next Generation Modelling Systems Programme is set to revolutionise how we simulate and predict the Earth’s atmosphere, oceans, and environment.
This ambitious initiative is not just about harnessing greater computational power, it’s about reimagining the very foundations of our modelling systems to meet the challenges and opportunities of future technologies.
Why next generation modelling systems are needed
Supercomputers have driven advances in weather and climate prediction since the 1950s, with exponential growth in processing power. However, physical and engineering constraints mean that further progress will require not only more processors but also new types of processors, particularly heterogeneous architectures that combine different processor types. To fully exploit these new architectures, the design of our modelling systems must change, both in the scientific algorithms that emulate the physics of the atmosphere and oceans, and in the software infrastructures that encode them.
Designing systems for architectures that do not yet exist is a formidable challenge. Flexibility and portability are essential, allowing software to run efficiently on diverse hardware without needing to be rewritten for each new processor type. Modern software design principles, such as separation of concerns, help by isolating hardware-specific aspects from science-specific aspects. This approach requires close collaboration between computational scientists, software engineers, and weather and climate scientists, but the rewards are substantial: more usable, efficient, and future-proof systems.
Programme aims and strategic importance
The Met Office’s Next Generation Modelling Systems Programme aims to reformulate and redesign its entire suite of weather and climate research and operational systems. This includes oceans and the broader environment, ensuring that the Met Office and its partners can fully exploit future generations of supercomputers for the benefit of society. The programme is a Corporate Strategic Action and a central theme of the Met Office’s Research and Innovation Strategy for the coming decade.
Programme timeline
Implementation of the Next Generation Modelling Systems for operational Numerical Weather Prediction (NWP) will be delivered towards the end of the phased implementation of the Met Office’s new supercomputer to be completed in 2027. The timeline includes phased operational implementation, parallel suites, and a four-year model evaluation period, with interim delivery of new capabilities and prototype systems.
Improving forecast and prediction accuracy
Harnessing greater computational resources will enable models to represent the atmosphere and oceans with finer granularity, improving the accuracy of key physical processes, better representing uncertainty, and capturing extreme events further ahead in time. Modern software practices will also enhance usability, robustness, and flexibility, improving the user experience and enabling new partnerships to deliver more science, more quickly.
Programme Overview: Driving Next-Generation Modelling
Our programme spans the full modelling workflow, from processing observations to visualising and verifying outputs. Here’s what’s been happening:
Observation Processing & Data Assimilation
We’re replacing the current OPS and VAR systems with a modern framework built on JEDI (Joint Effort for Data assimilation Integration). This modular system connects models, observations, and algorithms through a standard interface. Two key themes lead the way: JOPA – Observation processing and JADA – Data assimilation.
GungHo Atmospheric Science Project (GHASP)
The GHASP project has delivered cutting-edge science within the LFRic infrastructure, improving accuracy for global and regional forecasts. A major innovation? Moving from latitude–longitude grids to cubed-sphere grids, ensuring uniform point distribution.
LFRic & PSyclone
Born from GungHo, LFRic provides next-gen software infrastructure for atmospheric modelling. It separates science algorithms from parallel code for scalability across supercomputers. PSyclone automates and optimises parallel code, boosting performance and portability. Supporting projects include LFRic Inputs, which streamlines data flow and file handling.
Marine Systems
Our ocean and wave models - NEMO, NEMOVAR, and WAVEWATCH III—are adapting for future architectures. Work focuses on cost-effective CPU improvements, GPU adaptation, and strategies for emerging systems, using a “separation of concerns” approach.
Atmospheric Composition
We’re porting the UKCA (UK Chemistry and Aerosol) sub-model to LFRic, integrating gas-phase chemistry and aerosol modelling via GungHo.
Coupled Earth System Modelling
A new model links atmosphere, ocean, ice, land, and hydrology using: GHASP for atmosphere, NEMO for ocean and TRIP for hydrology. Coupling is handled by OASIS3-MCT, enabling efficient data exchange and realistic simulations. This modular design ensures scalability and future integration.
Verification System
Replacing VER, the new system uses MET and METplus (open-source tools from NCAR) to verify atmosphere and ocean models.
Research to Operations (NG-R2O)
This project integrates all Next Generation components into operational NWP systems, ready for the next supercomputer by 2027/28. It covers global, UK, regional, and sub-km models.
Research to Climate Use
Working with other projects, this initiative builds climate workflows to exploit new supercomputing capabilities.
Atmospheric Dispersion
The NAME system continues to deliver operational services while exploring coupling with LFRic for simultaneous operation.
Training & Usability
Helping Unified Model users transition smoothly, this project develops training resources and improves user experience for long-term sustainability.
Integration
Provides critical path analysis, technical capabilities, and interfaces for downstream customers. It supports multiple projects with model I/O, software stacks, and cloud deployment.
Model Evaluation
Regional LFRic Evaluation
Establishes a robust baseline for research and Global LFRic Evaluation – Builds and assesses a global coupled model within LFRic.
Model Optimisation
Improves LFRic performance through PSyclone enhancements, GPU infrastructure, and algorithmic optimisations.
Visualisation & Tools
Extended the Iris Python package for advanced visualisation and analysis of LFRic’s unstructured UGRID data.
Fab Build System
An open-source build tool tailored for LFRic’s design, supporting conventional and novel workflows.
Working in partnership
The success of the Next Generation Modelling Systems Programme depends on collaboration with a wide range of partners including operational and research centres in the UM Partnership, such as the Bureau of Meteorology (Australia), the National Institute of Water and Atmospheric Research (New Zealand), now Earth Sciences New Zealand, and Oak Ridge National Laboratory (US Air Force). The Joint Centre for Satellite Data Assimilation and the National Centre for Atmospheric Research, have contributed to the development of JEDI and METplus. The Science and Technology Facilities Council’s Hartree Centre has played a key role in developing PSyclone, reinvigorating long-term collaborations.
Looking ahead
The Next Generation Modelling Systems Programme is a bold step towards strengthening the long-term resilience the Met Office’s weather and climate prediction capabilities. By embracing new supercomputing architectures, modern software practices, and collaborative innovation, the Met Office is ensuring that its forecasts and climate predictions remain at the cutting edge - delivering greater accuracy, usability, and societal benefit for years to come.
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.
