LFRic advances weather and climate modelling through supercomputing innovation, collaboration, and next-generation training programs.
Why weather and climate modelling must evolve
From severe weather forecasting to understanding climate change, weather and climate models are at the heart of modern environmental science. As scientific questions become more complex and computing technology evolves, the tools behind those models need to evolve too, taking advantage of increasingly-powerful supercomputers while remaining flexible, maintainable, and scientifically robust.
What is LFRic?
To meet this challenge, the Met Office in collaboration with the Momentum Partnership, Natural Environment Research Council (NERC) and the Science and Technology Facilities Council (STFC), have developed LFRic a next-generation modelling system designed to exploit advances in supercomputing and support the future of weather prediction and climate projections.
The Momentum Partnership brings together operational and research organisations from around the world to develop and use Momentum®, a seamless modelling framework that underpins weather and climate science and services. Through shared science, software and infrastructure, the Partnership helps to enable effective collaboration, accelerate innovation, and support improvements in weather prediction and climate projections.
By advancing science, creating modern modelling tools, and supporting training and skills development, the Partnership is helping build a community of researchers, developers and operational users ready to work with the next-generation of weather and climate models.
Developing new modelling systems is not just about software. It also requires an international community of researchers, developers and operational users who can work together, share expertise and build new skills. A key part of this effort is the launch of the new LFRic Atmosphere Training course, helping researchers and users build the skills and confidence needed to work with the LFRic ecosystem and take advantage of emerging computing technologies.
Nigel Wood, Principal Fellow for Next Generation Modelling Systems said: "As a key element of Momentum, LFRic is an important step towards the future of weather and climate modelling, helping ensure that our science can take advantage of emerging computing technologies while remaining accessible to the wider research community.
"One of the many strengths of the Momentum Partnership is the way it brings together expertise from across the weather and climate community. I am excited to see the launch of the new LFRic Atmosphere Training course as that will help turn that shared expertise into practical learning, supporting the next generation of researchers and operational users working with LFRic."
How LFRic is designed for future science and computing
LFRic has been designed to be able to better exploit the vast number of processors of current and future high-performance supercomputers, as well as the ever-growing range of different processor technologies. As such, LFRic provides a flexible foundation for the efficient running of the next generation of weather and climate models.
Named after pioneering weather forecaster Lewis Fry Richardson, LFRic introduces three major developments: a new way to solve the equations; a more scalable way of representing the globe; and new software that helps the model run efficiently on different types of supercomputers.
New ways of solving atmospheric equations
At the heart of every weather and climate model is the mathematical engine that calculates how air moves and transports heat around the planet. This is known as the dynamical core.
LFRic introduces a new dynamical core called GungHo. Built with modern supercomputers in mind, GungHo helps scientists run larger and more detailed simulations while maintaining the accuracy needed for operational weather prediction and climate projections.
Fig. 1: Representation of Gungho, the dynamical core of LFRic modelling system. Gunho solves the equations that describe how the atmosphere move.
New approaches to representing the globe
Many weather and climate models traditionally represented the Earth using a regular grid defined by lines of latitude and longitude. This approach has served the scientific community well for decades, but it can become less efficient on modern supercomputers and creates challenges near the poles.
LFRic takes a different approach. In simple terms, this means dividing the globe into a cube-like grid known as a cubed-sphere mesh which creates a more even representation of the Earth's surface and avoids some of the inefficiencies of a traditional latitude-longitude grid. This new mesh scales more efficiently on large computing systems, supporting increasingly high-resolution modelling of weather and climate processes.
Fig 2. Representation of a traditional latitude-longitude grid and the cubed-sphere mesh which is a more scalable way of representing the Earth's surface.
New software tools that help models run efficiently on modern supercomputers
Modern supercomputers are becoming increasingly diverse, using different architectures and processor types such as CPUs and GPUs to deliver performance.
Traditionally, scientists often needed to spend considerable effort to implement simulated scientific processes into new machines. LFRic addresses this challenge using PSyclone, a software framework developed by the STFC that separates scientific code from machine-specific optimisations. This allows scientists to focus on the science, and software engineers to focus on the technical details needed to run the model efficiently. The result is a system that can incorporate cutting-edge science while remaining more flexible, easier to maintain and ready for future supercomputers.
Fig 3. Representation of PSyclone software that separates scientific code from machine-specific optimisations.
Introducing the LFRic Atmosphere Training course
Together, these innovations are helping transform the way weather and climate models are developed and run. However, adopting a new modelling system also means learning new tools, workflows, and ways of working.
The Momentum Partnership team, in collaboration with the University of Exeter and National Centre for Atmospheric Science (NCAS), has launched the LFRic Atmosphere Training course. It features a practical, self-paced introduction to the LFRic ecosystem to help researchers and users build confidence with this new system.
The training has been designed for both newcomers and experienced users, the training provides a foundation for understanding, configuring, and running LFRic Atmosphere simulations.
The course covers four key areas:
- An introduction to LFRic and the Momentum framework.
- Working with unstructured grids and data.
- Understanding the LFRic software architecture, including the GungHo dynamical core and PSyKAl approach associated with PSyclone, which helps models run efficiently across different supercomputing architectures.
- Running and managing global, regional, and idealised science configurations.
Alongside the theory, users gain hands-on experience building applications, running simulations on HPC platforms, and analysing model output using Python and Jupyter notebooks.
Building an international community around LFRic
The training materials are hosted openly on GitHub and are intended to grow with the LFRic community. We welcome feedback, issue reports, and contributions from users and developers.
We are also working with platform support teams across partner organisations to make the hands-on exercises available on a wider range of supercomputing systems, helping ensure that researchers can access the training wherever they work.
How to get started
Whether you are taking your first steps with LFRic or preparing for the future of weather and climate modelling, the LFRic Atmosphere Training course provides a practical route into the system.
Start your LFRic training journey.
